Read Anytime Anywhere Easy Ebook Downloads at ebookmeta.

com

Computer Vision and Augmented Reality in iOS:

OpenCV and ARKit Applications 1st Edition Ahmed
Fathi Bekhit

https://ebookmeta.com/product/computer-vision-and-augmented-
reality-in-ios-opencv-and-arkit-applications-1st-edition-
ahmed-fathi-bekhit/

OR CLICK HERE

DOWLOAD EBOOK

Visit and Get More Ebook Downloads Instantly at https://ebookmeta.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

NET Developer s Guide to Augmented Reality in iOS Building

Immersive Apps Using Xamarin ARKit and C 1st Edition Lee
Englestone
https://ebookmeta.com/product/net-developer-s-guide-to-augmented-
reality-in-ios-building-immersive-apps-using-xamarin-arkit-and-c-1st-
edition-lee-englestone/
ebookmeta.com

Mastering ARKit: Apple’s Augmented Reality App Development

Platform 1st Edition Jayven Nhan

https://ebookmeta.com/product/mastering-arkit-apples-augmented-
reality-app-development-platform-1st-edition-jayven-nhan/

ebookmeta.com

Mastering ARKit: Apple’s Augmented Reality App Development

Platform 1st Edition Jayven Nhan

https://ebookmeta.com/product/mastering-arkit-apples-augmented-
reality-app-development-platform-1st-edition-jayven-nhan-2/

ebookmeta.com

The World Formula: A Late Recognition of David Hilbert‘s

Stroke of Genius 1st Edition Norbert Schwarzer

https://ebookmeta.com/product/the-world-formula-a-late-recognition-of-
david-hilberts-stroke-of-genius-1st-edition-norbert-schwarzer/

ebookmeta.com
Target C#: Simple Hands-On Programming with Visual Studio
2022 1st Edition Gerard Byrne

https://ebookmeta.com/product/target-c-simple-hands-on-programming-
with-visual-studio-2022-1st-edition-gerard-byrne/

ebookmeta.com

GitOps and Kubernetes Continuous Deployment with Argo CD

Jenkins X and Flux 1st Edition Billy Yuen Alexander
Matyushentsev Jesse Suen Todd Ekenstam
https://ebookmeta.com/product/gitops-and-kubernetes-continuous-
deployment-with-argo-cd-jenkins-x-and-flux-1st-edition-billy-yuen-
alexander-matyushentsev-jesse-suen-todd-ekenstam/
ebookmeta.com

Rushing Into Love A small town sports romance 1st Edition

Kara Kendrick

https://ebookmeta.com/product/rushing-into-love-a-small-town-sports-
romance-1st-edition-kara-kendrick/

ebookmeta.com

The Misdiagnosis Casebook in Clinical Medicine: A Case-

Based Guide 1st Edition Hassaan Tohid

https://ebookmeta.com/product/the-misdiagnosis-casebook-in-clinical-
medicine-a-case-based-guide-1st-edition-hassaan-tohid/

ebookmeta.com

Stochastic Numerics for Mathematical Physics 2nd Edition

Grigori N. Milstein

https://ebookmeta.com/product/stochastic-numerics-for-mathematical-
physics-2nd-edition-grigori-n-milstein/

ebookmeta.com
Concrete Algebra With a View Toward Abstract Algebra
Benjamin Mckay

https://ebookmeta.com/product/concrete-algebra-with-a-view-toward-
abstract-algebra-benjamin-mckay/

ebookmeta.com
Ahmed Fathi Bekhit

Computer Vision and Augmented

Reality in iOS
OpenCV and ARKit Applications
Ahmed Fathi Bekhit
Boca Raton, FL, USA

ISBN 978-1-4842-7461-3 e-ISBN 978-1-4842-7462-0

https://doi.org/10.1007/978-1-4842-7462-0

© Ahmed Fathi Bekhit 2022

This work is subject to copyright. All rights are reserved by the

Publisher, whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other
physical way, and transmission or information storage and retrieval,
electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.

The use of general descriptive names, registered names, trademarks,

service marks, etc. in this publication does not imply, even in the
absence of a specific statement, that such names are exempt from the
relevant protective laws and regulations and therefore free for general
use.

The publisher, the authors and the editors are safe to assume that the
advice and information in this book are believed to be true and accurate
at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, expressed or implied, with respect to the
material contained herein or for any errors or omissions that may have
been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Apress imprint is published by the registered company APress

Media, LLC part of Springer Nature.
The registered company address is: 1 New York Plaza, New York, NY
10004, U.S.A.
This book is dedicated to the memory of Daniel Bakhtiyarov.
Acknowledgments
I would like to acknowledge the extraordinary support I have received
from family and friends throughout the journey of writing this book; I
am extremely fortunate and grateful to have such an amazing support
system around me. Thank you, David Parshenkov, Fathi Bekhit,
Mohamed Bekhit, Hazem Abdeltawab, Youssef Gamal, Ahmed
Abdulkareem, Seif Hediya, and Amr Aly, for all the love and support you
have given me that encouraged me throughout this journey. I love you
all.
This book is a reality today thanks to Aaron Black, Senior
Acquisitions Editor at Apress, for believing in me; Jessica Vakili,
Coordinating Editor at Apress, for the continual support and feedback;
Akanksha Devkar, Deep Learning Engineer, for the amazing feedback to
make this book comprehensive; and Affan Abbas, Deep Learning
Researcher, for the brilliant input and feedback on state-of-the-art
computer vision technologies used in academia today.
Last but not least, thank you, Leo, for being there for me in the late
nights I spent writing this book. I could not have done it without you.
Table of Contents
Chapter 1:​Introduction to Computer Vision
What Is Computer Vision?​
An Overview on How Computer Vision Works
Why Do We Need Computer Vision?​
The Evolution of Computer Vision
Traditional vs.​Deep Learning–Based Computer Vision
Computer Vision in Augmented Reality
Summary
Chapter 2:​Introduction to Augmented Reality
What Is Augmented Reality?​
Different Types of Augmented Reality
Summary
Chapter 3:​Image and Video Processing Fundamentals
Mathematics Overview
Image Processing Basics
Standard Algorithms
Object Detection
Summary
Chapter 4:​Computer Vision Applications
Overview
Applications of Computer Vision
Read, Write, and Display Images
Change the Color Space
Resize the Images
Rotate the Image
 Bitwise Operation
Detect the Edges
Image Filtering
Detect Face
Computer Vision Algorithms
Image Classification in Deep Learning
Object Detection
Object Detection Types
Importance of Object Detection
Detection of Anomaly
Object Tracking
Semantic Segmentation
Vehicles That Are Automatic
Medical Diagnostic Picture
Instance Segmentation
Requirements
Work on Image Reconstruction
Implementing Vision Applications
A) Face Detection/​Recognition
B) Body Detection/​Tracking
C) Image-to-Text Analysis
Chapter 5:​Custom Object Tracking
Overview
Real-Time Object Tracking
Background
Utilizing Code
 Libraries
Conclusions
Points of Concern
Custom Barcode Generator/​Reader
Summary
Chapter 6:​Augmented Reality Using OpenCV
Overview
Marker-Based AR
Location-Based AR
Marker-Based Augmented Reality
Summary
Chapter 7:​Augmented Reality Using ARKit and OpenCV
Why Do You Use OpenCV instead of Core ML?​
Multiple Platforms Are Supported
ARKit and OpenCV Integration
To Sum Up
Prerequisite
Hardware
Software
Body Tracking Step by Step
Step 1:​Set Up the Primary Scenario
Step 2:​Set Up the Skeleton
Step 3:​Locate the Followed Bodies
Step 4:​Show the Skeleton
Step 5:​Create and Deploy
Face Tracking in ARKit 3
 User Level:​Basic
Setup and Specifications for ARKit 3
Session Objects in Augmented Reality
Prefab AR Glasses
ARGlassesControl​ler Class
ARFace Data and Face Detection in ARKit 3
Color Selector
Summary
Index
About the Author
Ahmed Fathi Bekhit
is Cofounder and Chief Technology
Officer at Magic Studio, a software
company based in Boca Raton, Florida,
that focuses on developing cutting-edge
video processing and streaming
technologies. In the early days of
Ahmed’s career in software engineering,
he was awarded the Apple WWDC
Scholarship three times in a row and was
also titled “Apple’s Whiz Kid” by The Wall
Street Journal at the age of 15.
As of today, Ahmed has worked as a
software engineer with a few startups
for seven years, and he has been writing software for over a decade. He
also regularly contributes to the open source community and has been
known for creating ARVideoKit, an iOS framework that renders ARKit
content to capture videos, GIFs, and live photos with Augmented Reality
components. In addition to developing software, Ahmed has interest in
writing educational material that is focused on software engineering to
share his background and knowledge with others. For example, he
contributed various articles on the tutorial site AppCoda. He also
published educational material through Stanford University’s Scholar
Initiative. Additionally, over the past few years, Ahmed developed and
published independent iOS applications that reached over five million
users. Ahmed’s engineering work has been notably integrated in
Emmy-winning software and licensed to help with the COVID vaccine
manufacturing pipeline.
About the Technical Reviewer
Massimo Nardone
has more than 22 years of experience in
security, web/mobile development, and
cloud and IT architecture. His true IT
passions are security and Android.
He has been programming and
teaching how to program with Android,
Perl, PHP, Java, VB, Python, C/C++, and
MySQL for more than 20 years.
He holds a Master of Science degree
in computing science from the University
of Salerno, Italy.
He has worked as a project manager,
software engineer, research engineer,
chief security architect, information
security manager, PCI/SCADA auditor, and senior lead IT
security/cloud/SCADA architect for many years.
© Ahmed Fathi Bekhit 2022
A. F. Bekhit, Computer Vision and Augmented Reality in iOS
https://doi.org/10.1007/978-1-4842-7462-0_1

1. Introduction to Computer Vision

Ahmed Fathi Bekhit1
(1) Boca Raton, FL, USA

This chapter will focus on what computer vision is, why we need it, the
evolution of the technology, its different applications, and how it is used
in Augmented Reality.

What Is Computer Vision?

Vision is the ability to analyze and interpret scenes and objects of
interest. Human vision has been studied for hundreds of years to
understand how the visual process works. The human visual process is
one of the most complex processes to understand. In fact, to this day,
vision scientists have not yet found a complete answer to how the
visual process works. However, vision scientists’ discoveries on how the
human visual process begins and a little beyond that inspired computer
scientists to develop what we know today as Computer Vision. Vision
researchers and scientists describe that the visual process begins with
the eyes processing signals of light and converting them into scenes and
images for the brain’s visual cortex to analyze and interpret. A
breakthrough in vision research in the 1950s discovered that the visual
process begins by detecting the simple structures and edges of an
image to help build up a more detailed interpretation as the visual
information becomes more complex. The breakthrough vision research
inspired computer scientists to develop the preprocessing Computer
Vision algorithms we use today to initiate every computer vision task.
Compared to a typical computer today, the human brain computing
speed is significantly slower than a computer’s computing speed, yet
the human brain performs vision tasks much faster and significantly
better than any computer. Hence, researchers’ inspiration to develop
Computer Vision algorithms has always been the evolution of vision in
nature.
Computer Vision is the field of studying and developing technology
that enables computers to process, analyze, and interpret digital
images. Today, Computer Vision applications can be found in several
industries, such as industrial robots, medical imaging, surveillance, and
many more. All these applications have one principal mission, and it is
processing, analyzing, and interpreting the contents of digital images to
perform a task relevant to an industry’s needs, which will be referred to
as a vision task in the rest of this book. A vision task is any kind of task
that requires processing, analyzing, or interpreting the contents of
digital images and videos. For reference, a video is a sequence of digital
images, typically consisting of 30–60 digital images per second, also
referred to as frames.
Computers display digital images very often. When a digital image is
given to a computer as an input, the computer reads it as a two-
dimensional array of pixels; it can also be defined as a two-dimensional
matrix. An image matrix consists of M columns and N rows. The size of
an image in pixels can be determined by finding the product of M
columns and N rows (M × N), where M is the width and N is the height of
the image. A pixel position is identified by its x and y coordinates (x, y)
in the matrix. The coordinate system in computer graphics and digital
images is slightly different from a typical Cartesian coordinate system;
the point of origin (0, 0) in a digital image begins from the top-left
corner of the image. Therefore, x is increasing from left to right, and y is
increasing from top to bottom (see Figure 1-1).
Figure 1-1 Digital Image Coordinate System
In a grayscale digital image, a pixel value can be represented in
various formats; however, the most common format used is an 8-bit (1-
byte) unsigned integer ranging from 0 to 255 where 0 is black and 255
is white. Any value between 0 and 255 defines a different shade of gray
(see Figure 1-2).

Figure 1-2 Grayscale Image Converted to a Matrix

Figure 1-2 demonstrates a 16 × 16 grayscale image converted to

pixel values that are represented in 8-bit unsigned integers.
In a color digital image, a pixel value must be specified in three
colors – red, green, and blue – which represents an RGB color space.
Thus, the pixel value is represented as a vector of three numbers where
each number ranges from 0 to 255. As the number increases, the
brightness of each color increases, also referred to as color intensity .
For example, if a given pixel value is (255, 255, 255), this means the red,
green, and blue intensities (or brightness) are set to their highest value,
which defines a white pixel (see Figure 1-3).
Figure 1-3 RGB Color Space and Color Intensity Example
In Figure 1-3 you can see how the color intensity can affect the final
resulting color. The higher the color intensity, the brighter the resulting
color.
Although rendering a digital image appears to be a computer-
friendly task, computer vision requires more complex processing to
interpret, analyze, or manipulate the contents of a digital image.
Computer Vision consists of two core applications: manipulating
digital images and processing digital images to perform vision tasks.
Manipulating digital images includes image enhancement, restoration,
reconstruction, and compression. Image enhancement and restoration
are typically used to improve the quality of distorted images.
Reconstruction of images takes advantage of a set of two-dimensional
images from different angles to reconstruct a three-dimensional image.
However, some implementations that use deep learning attempt to
reconstruct a three-dimensional image from a single two-dimensional
image. Image compression is used to convert images into a more
efficient representation to minimize storage usage. Processing digital
images, on the other hand, may use some image manipulation
techniques to perform vision tasks that match, describe, or recognize
the contents of an image. The process of matching and describing
contents of an image mainly depends on segmenting images into
smaller parts and comparing the segmented parts to other template
images and measuring the relationships between the segmented parts
and the template images (see Figure 1-4).
Image Credits: Racool_studio/Freepik.comFigure 1-4 Template Matching Abstract
Example
Figure 1-4 demonstrates a sample result of a template matching
algorithm.

An Overview on How Computer Vision Works

In order for a computer to perform a vision task, such as face detection,
it requires performing several smaller tasks to reach the target task. For
example, in order to perform a vision task, a computer typically has to
perform various digital image preprocessing techniques, such as
grayscale manipulation, edge enhancement and detection, noise
removal, image restoration, interpolation, and image segmentation.
Image preprocessing is the first stage in every vision task, and it plays a
significant role to have a successful outcome. The purpose of
preprocessing is extracting the valuable parts in a given image and
getting rid of unwanted distortions. The valuable parts help in
developing useful descriptions of pixels, shapes, and surfaces in a given
image. As we mentioned earlier, a pixel value is defined by either a 1-
byte integer (when it is in a grayscale image) or a vector of integers
representing an RGB color space (when it is in a color image); and each
pixel value has a different intensity level representing the brightness of
the color. The changes of intensity in a given image provide clues about
the structure of shapes and surfaces in the image.
In the preprocessing stage, we try to extract this information
without any knowledge of the contents of an image using techniques,
such as noise removal, edge enhancement, and image segmentation.
Noise removal is the process of smoothing an image to reduce and
eliminate unwanted noise/distortion. However, sometimes noise
removal techniques end up blurring the image; in this case, edge
enhancement techniques come in handy. Edge enhancement helps with
improving the sharpness of the edges in an image by increasing the
intensity level difference between the boundaries of two regions (see
Figure 1-5).

Figure 1-5 Edge Enhancement/Detection Example

The image on the left is the original image. The image on the right is
the resulting image after applying an edge enhancement technique. As
you may have noticed, the right image appears to “outline” the edges of
the dog, which can be used in various vision tasks, such as image
classification.
Image segmentation is the process of separating an image into
smaller parts called segments. Segments are used to process the parts
of an image separately, because in many cases an image has more than
just one unique object or feature that can help identify the contents of
the image. Image segmentation is considered an intermediate step
between preprocessing and processing an image for object detection or
segment classification (see Figure 1-6).
Figure 1-6 Abstract Computer Vision Pipeline
After preprocessing a digital image, the processing steps begin, and
in many cases, it may involve techniques such as feature extraction,
texture analysis, and pattern recognition. Depending on each vision
task, you would pick the more relevant technique to perform the task.
For example, to perform an image classification task, you would use a
feature extraction technique to extract unique features in the segments
of a given image. The segments of an image are extracted during the
image segmentation process. Once the unique features are extracted
from the image segments, they are used to find a distinct pattern
matching predefined or labeled image segments; a labeled image
segment is a piece of an image that is manually defined under human
supervision (Figure 1-7).

Figure 1-7 Labeled Image Segments Example

The process of finding an exact match to a labeled segment is called

feature matching; and the process of finding a distinct pattern is called
pattern recognition. Figure 1-6 demonstrates an abstract Computer
Vision pipeline and what a given image has to typically go through
when a computer is performing a vision task. Further details on
different image processing techniques will be covered throughout the
book.

Why Do We Need Computer Vision?

Since the Internet has been accessible to billions of users around the
world, it has enabled many to capture, produce, and publish media-rich
content, such as images, videos, and animated images (also known as
GIFs). User-generated content defines the Internet we know today, and
the amount of media-rich content published per day is rapidly
increasing. Today, YouTube publishers alone contribute 300 hours of
video per minute; that is 3 million hours of video per week. There is no
human who can handle watching this much video at this rate. YouTube
and many similar companies that allow user-generated content on their
platforms are challenged to index, classify, and categorize billions of
user-generated images and videos to provide their viewers the most
relevant content they would watch. The categorization of all this
content cannot be properly handled by a typical software; most of the
existing approaches depend on the publishers’ text input, such as video
title, description, keywords, etc. However, in many cases, the
publisher's text input can be irrelevant; and their content may contain
inappropriate or violent scenes that should not be displayed to certain
age groups, or many viewers would want to opt out of viewing such
content. The only technology that would enable large-scale image and
video indexing, classification, and categorization is Computer Vision.
Computer Vision is needed by companies, such as YouTube, to index
and categorize the massive amount of media-rich content. In addition
to categorizing images and videos, computer vision can be used to
– Provide a visual description of images and videos to the visually
impaired users and customers.
– Filter inappropriate content to certain age groups.
– Warn viewers of the type of scenes and content included in an image
or a video.
– Detect manipulated images and videos that may cause harm to the
viewers or persons in the manipulated content.
In addition to the Internet-related problems Computer Vision can
solve, there are many other problems across various industries that
need computer vision technology to be solved. Now, you can find
computer vision technology integrated in many things we interact with
on a daily basis, such as cashier systems, surveillance systems,
biometric scanners, medical equipment, transportation, and many
more.
– Cashier systems integrate computer vision technology to enable self-
checkout kiosks.
– Surveillance systems use image and video enhancement algorithms
to detect and analyze the contents of surveillance images and videos.
– Biometric scanners use edge detection and enhancement techniques
to collect and analyze biometric information (i.e. fingerprints, facial
features, eye features... etc).
– Medical equipment takes advantage of various computer vision
techniques, such as image reconstruction and restoration to retrieve
useful information from medical devices and scanners.
The list can go on for pages demonstrating all the different
applications of computer vision. In fact, the need of computer vision
technology has been only increasing over the past decade. Top
companies are adopting the technology in their products, and new
companies are starting to provide computer vision–powered products.
The impact of computer vision technology is limitless, and it will be
adopted by more industries and organizations around the world over
the next couple years. You may have already seen the initial iterations
of the future of computer vision applications, such as self-driving
vehicles, autonomous manufacturing, and education via Augmented
Reality.
In 2019, companies’ demand on Augmented Reality, Virtual Reality,
and Computer Vision engineers has increased by 1500% (State of
Software Engineers, Hired.com). Computer Vision will have a great
impact on society in the following years and will require more
contributors to the technology, which this book will intend to do by
enabling software engineers and persons with interest in computer
vision to be part of the development of this revolutionary technology.

The Evolution of Computer Vision

In 1959, a neurophysiology research conducted a couple experiments
on a cat’s visual system. The experiments placed electrodes in the main
visual cortex of a cat’s brain, and the researchers observed the neural
response of the visual cortex as they presented various images to the
cat (Figure 1-8).

Figure 1-8 Hubel and Wiesel’s Experiment

As the experiments went on, the researchers noticed a pattern in
the cat’s neural response when presented with a new image. The
research concluded that there are simple and complex neurons in the
visual cortex. And the visual process begins by detecting the simple
structures and edges of an image to help build up a more detailed
interpretation. As the visual information becomes more complex, the
visual cortex depends on the complex neurons to interpret the
remaining details of an image. This research later became to be one of
the most influential research in the field of Computer Vision. It inspired
Computer Vision researchers to achieve what is known today as image
preprocessing.
In July of 1966, Seymour Papert, a professor at the Artificial
Intelligence Lab at Massachusetts Institute of Technology (MIT),
initiated “The Summer Vision Project.” The project was assigned to
undergraduate MIT students to develop pattern recognition technology
in an attempt to enable machines to see. The project was a very
optimistic plan to develop what we know today as Computer Vision in
one summer. There were various attempts in the 1960s to develop a
computer visual system; however, researchers eventually came to
realize that such a project would require years, if not decades, to
achieve. And here we are, 50 years later, and Computer Vision is still in
the research and development phases.
Since the 1960s, Computer Vision has evolved to the technology we
see in many applications today. One of the pioneering research in
computer vision was in the 1970s, where the research focused on
object recognition. The initial iteration of object recognition was
abstract and simple to prove the concept of object recognition. The
implementation used the line joints of simple objects, such as cylinders
and cubes, to identify them. In the following decades, researchers and
engineers developed the core techniques of traditional computer
vision, which consists of two principal stages:
1. Image preprocessing

2. Image processing

The image preprocessing stage focuses on enhancing image colors

and edges to help in detecting simple structures in a given image. The
processing stage uses the preprocessed image to perform one or more
of the following processing techniques:
– Image compression
– Image enhancement
– Image restoration and reconstruction
– Pattern matching and recognition
Although traditional image processing techniques such as the ones
listed previously have been used for decades since their development,
there have been a few problems these techniques alone would not
solve, such as recognizing objects in a significantly unusual or
unexpected context, such as handwritten words, digits, and images of
objects captured from an unusual angle (Figure 1-9).
Figure 1-9 Normal vs. Unusual Image of an Object
Consequently, in the 1980s, various groundbreaking research were
published. The research suggested using an artificial network of simple
and complex “neurons” that could recognize certain patterns. This
artificial network is today referred to as an Artificial Neural Network
(ANN). A Neural Network is a computing model that is designed to
recognize patterns in a given set of labeled data (labeled images in this
case). The network uses the given images to recognize a pattern and
predict the results of each image. Once the network predicts the results
of an image, it compares its prediction to the labels assigned to the
image. The comparison stage enables the network to “learn” from the
errors in its initial process, and then it performs the same process
multiple times until the network provides more accurate results. Neural
Networks often required lots of labeled data to provide significantly
good results. Hence, in the early days, a Neural Network was not used
to its full potential due to the lack of data.
In the early 2010s, a group of Computer Vision researchers initiated
a large-scale image dataset with millions of images and over 20,000
categories, called ImageNet. ImageNet was inspired by the need of
more data in the Computer Vision field. And it enabled the development
of state-of-the-art Neural Network Models that had a great influence on
the advanced computer vision applications seen today. Neural
Networks and their variations are considered part of the field of Deep
Learning, which is a subset of Machine Learning. Therefore, when
Neural Networks are used to perform vision tasks, I will be referring to
it as Deep Learning–based Computer Vision.

Traditional vs. Deep Learning–Based Computer

Vision
Traditional Computer Vision has existed ever since the computer vision
field began. Traditional Computer Vision techniques have been used for
decades to efficiently perform vision tasks. The traditional techniques
helped develop many algorithms that are commonly used today, such as
Scale-Invariant Feature Transform (SIFT) and Speeded-Up Robust
Features (SURF). Algorithms like SIFT and SURF have been a
fundamental part of developing today’s Augmented Reality technology
due to their efficient performance and accurate results. However, in
some cases, traditional computer vision techniques are not the best
choice due to the lack of accuracy. In these cases, the best option is
typically using a deep learning–based technique. Deep learning is used
in several ways to perform complex vision tasks, such as image
classification. Deep learning–based approaches typically outperform
traditional approaches when it comes to image classification. Image
classification is the process of classifying the contents of an image
between separate categories. For example, one of the most popular
image classification problems is classifying images of cats and dogs.
When a software is presented with an image of a cat, it should classify it
as a cat with high accuracy. Before deep learning, this task was nearly
impossible to achieve; however, deep learning pushed the limits of
what is possible to achieve in Computer Vision and successfully enabled
a computer to differentiate between a cat and a dog with high accuracy.
This would have not been possible without extremely fast computers
and datasets of millions of cat and dog images. You may wonder why
deep learning would require fast computers and a huge amount of data.
Well, the answer lies behind what deep learning is and how it really
works. Deep learning is a subset of machine learning that is largely
established on Artificial Neural Networks (ANNs), a computing model
inspired by how the human brain operates. Similar to the human brain,
an Artificial Neural Network consists of several computing “neurons.”
An artificial neuron in a Neural Network performs a simple computing
operation that communicates with other connected neurons in the
network to make a decision. Each neuron connection to another neuron
has a certain strength. The strengths of connections between neurons
are called weights (see Figure 1-10).

Figure 1-10 Feedforward Artificial Neural Network Model

Figure 1-10 is demonstrating a simple Neural Network consisting of
three layers—(1) Input Layer, (2) Hidden Layer, (3) Output Layer:
1. The Input Layer consists of multiple inputs where each input
represents a unique feature about the given dataset.

2. The Hidden Layer is a set of neurons that calculates and processes

in a certain way depending on the weight value.

3. The Output Layer is the last layer in a Neural Network that receives
the final results from the previous Hidden Layer.

A set of neurons represents a Neural Network Hidden Layer, and a

typical deep learning algorithm consists of multiple Neural Network
Hidden Layers. ANNs consist of two phases, the learning phase and the
performing phase. During the learning phase, a Neural Network takes
advantage of a labeled dataset to “learn” the correlation between the
data and their description (also known as labels). In our case, a labeled
dataset would consist of a set of images with a description of the
contents of each image. When a Neural Network is given a labeled
image, it attempts to process the image in the Hidden Layers and return
a prediction of the label as an output. Once an output is returned, the
Neural Network “grades” its result by comparing it to the original label
using a variety of functions called objective functions. If the grade does
not surpass a certain threshold, it will “learn” from it by adjusting the
weights of the connections between neurons in the Neural Network.
The Neural Network continues learning by presenting the same input
to itself, and the weights of the network are adjusted as it learns. This
process repeats multiple times until the Neural Network’s “grade”
surpasses the required threshold. The process of “learning” and
adjusting weights is called Backward Propagation of Errors, also known
as Backpropagation. To recap, here is a simple ordered list of the steps a
typical Artificial Neural Network goes through to “learn” something
new:
1. Receives labeled input

2. Attempts to predict the input’s label

3. Returns the predicted label as an output

4. Compares the predicted label to the actual label

5. Grades the results

6. Adjusts the strengths of connections between neurons, also known

as weights

7. Repeats the process until the final grade surpasses the threshold

Computer Vision in Augmented Reality

Augmented Reality (AR) is the technology that augments or displays
digital objects in the physical world through a camera live feed. In order
to display digital objects in the physical world, we would need to
understand and detect objects of the physical world in real time. In
some cases, the objects can be a specific template. In other cases, it can
be a horizontal surface, such as a desk and a floor, or it can be a vertical
surface, such as a wall and a door. To be able to detect objects and
surfaces, we would need to use Computer Vision techniques to display
digital objects in the physical world. There are three main types of
Augmented Reality: Marker-Based, Marker-Less, and Geolocation
Augmented Reality . Marker-Based AR requires a certain predefined
template to display digital objects on. For example, a page in a book or a
magazine can have a special icon so when a camera detects that icon it
would display a digital object mapped on the icon or the page (Figure 1-
11).

Figure 1-11 Marker-Based vs. Marker-Less AR

Unlike Marker-Based AR, Marker-Less AR does not require any

special predefined templates to display digital objects. Instead, it can
place digital objects on any horizontal or vertical surfaces using various
Computer Vision techniques. Geolocation AR, in many cases, depends
on either Marker-Based or Marker-Less AR and integrates location
sensor technologies to display digital objects in the physical world
based on visual and location data.
 Augmented Reality today is mostly accessible through mobile
devices, smartphones to be more specific. Although there have been
various attempts to integrate immersive AR experiences by developing
custom hardware and devices, such as Microsoft’s HoloLens, Magic
Leap Glasses, and Google Glasses, it is too early for consumers to adapt
to using new and unfamiliar hardware. Therefore, smartphones were
the perfect tool to introduce consumers to AR. In the early days of
mobile AR, using markers, such as a QR code or a template image, was
the main way to augment digital objects in the physical world. To
achieve that, techniques such as pattern recognition and pose
estimation were required. Pattern recognition is a technique that
depends on a predefined set of data describing features to find and
analyze in a given image. Pose estimation uses the detected pattern to
simulate a three-dimensional pose based on the camera angle (see
Figure 1-12); this technique is used to enable seeing a three-
dimensional object from all angles based on the camera’s point of view.

Figure 1-12 Body Pose Estimation

 Marker-Based AR allowed any user with access to a smartphone to
interact and use the technology; however, the requirement of having a
physical marker was not scalable nor practical for daily use. Therefore
Marker-Less AR had more potential due to its ease of use and
accessibility. In the early days of Marker-Less AR development, dual
cameras were required to have a reconstruction of the physical world.

Summary
In this chapter, we read about the history of Computer Vision. Grayscale
and color images were compared. We also studied why we need
Computer Vision and its evaluation. We also saw a simplified
architecture of a Feedforward Artificial Neural Network. Finally, we
looked at Augmented Reality.
To conclude, computer vision is a field of Artificial Intelligence (AI)
that enables computers and systems to derive meaningful information
from digital images, videos, and other visual inputs and take actions or
make recommendations based on that information. If AI enables
computers to think, computer vision enables them to see, observe, and
understand.
Computer vision works much the same as human vision, except
humans have a head start. Human sight has the advantage of lifetimes
of context to train how to tell objects apart, how far away they are,
whether they are moving, and whether there is something wrong in an
image.
In the next chapter, we’ll take a look at Augmented Reality and
different types of Augmented Reality.
© Ahmed Fathi Bekhit 2022
A. F. Bekhit, Computer Vision and Augmented Reality in iOS
https://doi.org/10.1007/978-1-4842-7462-0_2

2. Introduction to Augmented Reality

Ahmed Fathi Bekhit1
(1) Boca Raton, FL, USA

This chapter will focus on what Augmented Reality is, the evolution of
the technology, and its different applications.

What Is Augmented Reality?

Over the decades, humans have always used some sort of technology to
visualize data, tell stories, educate students, and even be entertained.
Data visualization has been essential for many organizations and
individuals to better understand information and make decisions.
Storytelling plays an important role in getting an audience engaged in
an activity. Education is critical to enable today’s students to form their
future. Forms of entertainment can be video games, sports, and
concerts. All these are examples of activities that can be much more
interactive by augmenting digital objects in the physical world and
enabling consumers to interact with them. To successfully augment
digital objects in the physical world, we would have to project light into
custom shapes and display it on physical objects. However, this
approach would require additional hardware that is not commonly
used by most consumers, such as screen projectors and holographic
projectors. The fact that additional hardware is needed to “augment”
digital objects in the physical world made the technology less
accessible and hardly scalable due to the high price tag of the hardware
needed. As the development of computer vision arose, more
researchers and engineers worked together to develop various
techniques to augment digital objects in the physical world and enable
new ways to interact with computers, to develop what we know today
as Augmented Reality.
Augmented Reality is the field of developing interactive user
experiences by reconstructing and displaying interactive digital objects
in the physical world. In the process of developing Augmented Reality
technology and applications, we are required to take in consideration
three core concepts:
1. Physical world reconstruction and mapping techniques

2. Digital object projection techniques

3. New human-computer interaction guidelines

One of the early approaches to augment and interact with digital

objects in the physical world was a project developed by a computer
researcher, Myron Krueger, who used a camera’s live visual data and
projected a computer interface to interact with the digital computer
components in the physical world (Figure 2-1).

Figure 2-1 Videoplace – Beginning of AR

 Many other approaches also depended on a camera’s live visual data
for scene understanding; but the projection of the digital components
was always changing. Some attempts created headsets and glasses that
would project the digital components on a small screen that was
wearable. And others would stream the camera’s live visual data to a
computer screen and interact with the digital components through the
screen. As computer vision technology evolved and camera devices
with a live feed on a portable screen became more popular and
accessible to consumers, the development of the Augmented Reality
technology that we are familiar with today began. In addition to
developing computer vision and digital object projection techniques to
achieve Augmented Reality, making Augmented Reality human-friendly
and easy to use is crucial to enable the full potential of the technology.
Computer vision, hardware, and human-computer interaction
researchers and engineers have been working together over the years
to create the ultimate Augmented Reality experience, which led to
developing the most common form of Augmented Reality seen today in
mobile phones.
Today, smartphones, such as iOS and Android devices, are the main
devices that enable users to experience Augmented Reality, given that
the majority of smartphones have powerful cameras, sensors, and
geolocation hardware. By making use of information provided by the
hardware found in smartphones, we can provide an immersive and
entertaining experience to the users through Augmented Reality. The
hardware pieces in a smartphone that provide sufficient information to
achieve Augmented Reality are single-lens or multi-lens cameras,
accelerometers, gyroscopes, magnetometers, and geolocation sensors
(Figure 2-2).
Figure 2-2 Single-Lens Camera
Single-lens cameras provide visual data of a scene from one angle,
which helps in detecting and tracking simple features of a surface in a
scene. Multi-lens cameras, on the other hand, provide visual data from
two different angles, which helps in detecting and tracking the depth in
a scene, and depth tracking enables a more accurate real-world
distance measuring. Accelerometers, gyroscopes, and magnetometers
are all part of a unit called Inertial Measurement Unit (IMU), which
enables measuring a device’s acceleration and estimating its velocity,
distance, orientation, and gravitational force. Geolocation sensors
provide information about the device’s location and geographic
coordinates, which enables augmenting and “pinning” digital objects in
location-specific areas (Figure 2-3).
Figure 2-3 Geolocation Sensor Measurement
As mentioned previously, there are three core concepts we need to
consider when developing Augmented Reality technology, and the first
concept is physical world reconstruction and mapping, also referred to
as scene reconstruction. Scene reconstruction is a computer vision task
whose core objective is to reconstruct video frames into a three-
dimensional scene. This vision task is essential to enable placing digital
objects in a specific part of a room and identify an object’s position in
the physical world using a certain coordinate value retrieved from the
reconstructed three-dimensional scene.
To perform this vision task, we would need to retrieve video frames
from either a single-lens or a multi-lens camera. Video frames retrieved
from a single-lens camera will make it more difficult to reconstruct the
scene due to the lack of depth information; however, thanks to a
computer vision technique known as structure from motion (SfM),
scene reconstruction is possible to perform via a single-lens camera.
SfM is a technique that uses an array of two-dimensional images to
detect unique features in the array of images and reconstruct a three-
dimensional scene (Figure 2-4). Unlike single-lens cameras, preforming
scene reconstruction using images from multi-lens cameras is slightly
less challenging since we can estimate depth from two or more images
and reconstruct a three-dimensional scene using the depth data.

Figure 2-4 Structure from Motion Model

Although scene reconstruction using vision-only techniques could
be sufficient in some cases, in many cases visual data coming from
handheld mobile phones is distorted due to the excessive motion and
unpredictable lighting conditions, leading to a large error margin and in
some cases failing to achieve scene reconstruction. In this case, Inertial
Measurement Units (IMUs) that can be found in most mobile phones
today would be used alongside the visual data to result in the best
scene reconstruction possible. Information retrieved from IMU devices,
such as orientation and acceleration, is extremely helpful when the
visual data is distorted or delayed for any reason since the IMU data is
provided at a higher frequency than the visual data, which helps in
detecting any small or significant changes in the motion and
environment. Such IMU data is aligned with the visual data to compute
the initial velocity and the physical coordinates of the device.
Using IMU data alongside the visual data from a camera has been a
common approach used in the Computer Vision and Augmented Reality
fields to perform scene reconstruction and other physical world
measurement tasks. In fact, it is typically referred to as Visual-Inertial
Odometry, VIO for short. VIO was initially developed for use in aerial
robotics, such as drones, and vehicles to enable autonomous
applications. As the technology evolved from only depending on multi-
lens visual to single-lens visual data, its applications increased and
eventually enabled Augmented Reality on smartphones. The
Augmented Reality applications of VIO mainly revolve around “Marker-
Less AR,” which enables augmenting digital objects in the physical
world without the requirement of a “marker.” An AR marker is a
uniquely identified object or image that is used to augment digital
objects on; this can be a QR code, image in a book, or human/animal
body parts, such as face, hands, and feet. To achieve Marker-Based
Augmented Reality, various vision tasks are required, such as object
pose estimation, annotation, and landmark detection.

Different Types of Augmented Reality

There are a few types of Augmented Reality that are commonly used
today; that includes Marker-Based, Marker-Less, and Geolocation-
Based Augmented Reality (AR). All three types’ core objective is to
augment digital objects in the physical world; however, the techniques
used to achieve each type vary. Additionally, the applications of each
type have different purposes. For example, Marker-Based AR requires
previously defined image features that we can use to augment digital
objects on. A commonly seen use case of this is Face Filters. AR Face
Filters augment digital objects on or surrounding a detected face and
track the face movements to change the digital objects’ position
accordingly (Figure 2-5).
Figure 2-5 Face Detection and Landmark Estimation
To achieve AR Face Filters, we would need to perform computer
vision tasks, such as face detection and face landmark
detection/tracking . Face detection is a process of segmenting and
detecting a face from a given image. Face landmark detection is the
process of detecting face features and tracking changes, such as eye
blinking, mouth opening/closing, and facial expressions. By combining
both information about the face in real time, we would be able to
successfully augment digital objects on a face. In addition to using a
detected face as a marker, there are other approaches that depend on
various other markers, such as a simple QR code, a unique image, and
custom physical objects (i.e., bottle, cup, hat, etc.). The
processing techniques used to achieve other marker-based AR
approaches are derived from the computer vision concepts of feature
detection, tracking, and pose estimation. As we get closer to the
implementation process at the end of this book, we will cover a more
in-depth description of the algorithms used to achieve marker-based
AR.
 Marker-less AR is a robust technique that does not require any
previous knowledge of an environment; however, it requires certain
pieces of hardware to achieve it. In the early days of marker-less AR
development, the type of hardware required to achieve marker-less AR
was not commonly owned by consumers, such as stereo cameras and
advanced IMU devices. As research and development of AR increased,
we have seen various approaches to minimize the use of uncommon
hardware and instead depend on hardware that is widely accessible to
consumers, such as low-cost IMU devices and single-lens cameras,
commonly found in smartphones today. In the previous section, we
briefly covered how marker-less AR is achieved today in most mobile
phones using Visual-Inertial Odometry techniques, which make use of
information retrieved from both IMU devices and a camera to perform
scene reconstruction alongside other computer vision tasks that would
enable marker-less AR. Marker-less AR applications are limitless, and
there is a lot of space for more creative and impactful applications.
Today, marker-less AR can be used to measure physical objects and
simulate an experience onto the physical world. These applications
have had a great impact throughout many industries including
education, design, tourism, and many more. To summarize, marker-less
and marker-based AR are fundamentally the same but have completely
different applications due to the restrictions each technology has.
Despite the fact that these two AR approaches have extremely different
applications, they can be used simultaneously to achieve a more
complex Augmented Reality experience, which leads us to Geolocation-
Based AR.
Geolocation-Based AR integrates both marker-less and marker-
based AR with geographic location data to augment and “pin” digital
objects in a specific geolocation coordinate (Figure 2-6).
Figure 2-6 Geolocation-Based AR

Summary
In this chapter, we discussed some basic concepts related to Augmented
Reality. After that, different types of Augmented Reality were discussed.
AR can be defined as a system that incorporates three basic
features: a combination of real and virtual worlds, real-time interaction,
and accurate 3D registration of virtual and real objects.
The core components of any AR-based tool are the processor,
sensors, input devices, and mainly the display. The display could be a
smartphone, a handheld device, smart glasses, or a head-mounted
display (HMD). The input devices are either cameras or web cams.
Sensors include gyroscopes and accelerometers.
In the next chapter, we will take a look at image and video
processing fundamentals in detail.
Other documents randomly have
different content
 “On the contrary, I asked a very simple question with a view to
finding out how Meredith looked when you last saw him. If I bungled
my meaning you must not take offense,” replied Curtis.
Colonel Hull covered his anger with bluff heartiness, while
inwardly registering a score to settle with the surgeon at some
future date.
“Certainly, I’ll answer any questions,” he exclaimed, with a broad
smile. “But you must admit your meaning was a bit obscure—and
from a total stranger; well, we’ll let it go, eh, Belle?” with a sidelong
look at Mrs. Meredith. “What is it you wish to know?”
“When you last saw Meredith, was he agitated or his normal self?”
questioned Curtis.
“Oh, he was a bit excited,” Hull admitted, with an air of candor.
“He called at my office one day last week and got uneasy over stock
quotations. He had been dabbling in oil, against my advice.”
“And that was the last time you saw him?” At Curtis’ polite
persistency Hull’s color deepened, but he was saved reply.
“Dad!” Lucille tapped him on his shoulder. “Mother is waiting in the
hall. She isn’t feeling well,” turning to Mrs. Meredith, who had risen
also, “so don’t keep her waiting, Dad.”
“I’ll come at once.” Colonel Hull waited courteously for Curtis to
precede him. “I am told, little girl, that John left you a very
handsome fortune.”
“In a codicil to his will,” Mrs. Meredith replied for Lucille who, a
step or two ahead, had not caught her father’s remark.
“Unfortunately the codicil cannot be found.”
Colonel Hull stopped dead in his tracks and glared at Mrs.
Meredith.
“What’s that?” he demanded. “Do you mean the codicil has been
suppressed—stolen, if you like it better?” meeting Mrs. Meredith’s
stony look with angry eyes.
“Dad!” Lucille laid a restraining hand on his arm and pressed it
warningly. “Don’t excite yourself. You will alarm mother.”
 Mrs. Hull, who had been too nervous to keep still, stopped her
aimless wandering about the square hall and waited for their arrival.
Lucille, in advance of the others, turned to Curtis.
“Mother,” she said, “let me introduce Doctor David Curtis,” she
hesitated before adding, “Anne’s fiance.”
“I am very pleased to meet you.” Savoir faire was not Mrs. Hull’s
strong point, and that she was ill at ease was as apparent to Curtis,
sensitive of his surroundings, as it was to his companions. She shook
his hand listlessly, then dropped it and pulled her evening cloak up
about her shoulders.
“The taxi is at the door,” announced Colonel Hull. “Come, Claire.”
But she lingered a moment to address Mrs. Meredith.
“When will John be buried?” she asked in an undertone.
“We will hold funeral services to-morrow morning in the chapel at
Oak Hill,” responded Mrs. Meredith. “Only the family will be present.
I thought Sam Hollister had told you of the arrangements; he has
them in charge.”
“I haven’t seen Sam.” Mrs. Hull kissed Lucille warmly, and then
shook hands with Curtis before she moved toward the front door.
“Good night, doctor. Oh, Belle,” with a change of tone, “it does make
me feel so badly to come here and not find John. He was so genial,
so kind. Only the last time I talked with him about Julian’s career, he
said I was my husband’s lodestar.”
Mrs. Meredith did not answer in words. After administering a cold
kiss on Mrs. Hull’s flushed cheek, and with a wave of her hand to the
Colonel, she turned back to Curtis, who stood waiting near the
entrance to the library.
“Lodestar is good, only spell it ‘load’” she commented, caustically,
but keeping her voice lowered so that it would not reach the Hulls.
“John had quite a sense of humor.”
Curtis smiled. “Are you going upstairs, Mrs. Meredith?” he asked.
“Yes—and you?” pausing on the lower step.
 “I’ll smoke awhile in the library; it is only nine o’clock,” as the
clock chimed the hour. “Good night.”
“Good night,” she echoed, and continued up the staircase.
Curtis listened until her soft footfall faded away in the distance,
then turned thoughtfully and entered the library. The servants had
spent but scant time after the inquest in replacing the furniture in its
accustomed places, and Curtis found some difficulty in moving
about.
“Oh, do be careful,” exclaimed a soft voice to his right, and a hand
touched his. “This way. I,” her dignity sat quaintly upon her, “I am
Anne.”
“As if it could be any one else!” Curtis spoke with involuntary
fervor, and Anne laughed shyly, then recollection returned to her,
and her expression grew serious.
“I came downstairs hoping to find you,” she explained, her color
mounting. “When I heard Cousin Claire and Cousin Julian talking in
the drawing-room I came in here to wait until they left. I want,” she
hesitated, selecting her words carefully, “to speak of Uncle John’s
plan for—for our marriage.” The last words came with a rush, then
she paused, tongue-tied.
Curtis Came to her rescue. “I understand,” he began gravely. “We
will call the whole affair off. In other words,” striving to spare her
embarrassment, “I release you from your promise.”
She plucked nervously at her gown. “It is you who do not
understand,” she said. “I don’t wish to be released.”
Curtis raised his head. Had his ears played him false?
“You mean,” he asked slowly, “that you wish to go on with the
marriage ceremony?”
“Yes.” The affirmative was little more than a whisper.
“But,” it was his turn to hesitate, “it seems now that you are very
wealthy; it is not necessary to carry out the bargain your uncle
wished to force upon you.”
 She did not answer at once. “I gave my word to him,” she
murmured. “I cannot break faith with the dead.”
The ticking of the mantel clock was distinctly audible in the
silence. Suddenly she spoke again, a catch in her voice.
“You hesitate—you do not wish to—to marry me?” she asked.
The hot color mounted to his brow and then receded.
“I only hesitate on your account,” he said. “In marrying me you
will be tied to a blind man—a failure.”
She did not reply at once. Instead, Curtis heard her move
backward a few steps and then a slight click sounded as an electric
lamp was switched on. Anne turned and regarded Curtis gravely
under its direct rays. There was none too much flesh even yet on the
tall, straight figure, but the air of alertness and poise which had
formerly been characteristic had returned to him. His face still bore
traces of mental suffering, although its unyouthful sadness had been
effaced.
“Because it is a bargain,” Curtis’ voice startled her from her
contemplation of him, “I wish it to be a fair one. You are offering me
the wherewithal to live. I can offer you nothing—”
“Perhaps,” she broke in swiftly, “I crave your friendship, your aid.”
Curtis felt his heart skip a beat and then race on.
“I will do anything, anything for you,” he replied, a trifle
unsteadily. “And will gladly carry out your uncle’s plan.”
“Thank God!” she whispered.
The portières were thrust back suddenly and Mrs. Meredith stood
on the threshold, with Hollister behind her.
“You may go to your room, Anne,” she said in icy tones.
A second later the portières dropped back into place and Curtis
was alone.
 CHAPTER IX

TWO PIECES OF STRING

David Curtis felt around his empty cigarette case and sighed
regretfully; he had not realized his rapid consumption of its contents.
The cigarettes had, at least, provided diversion of a sort. Since
Anne’s peremptory summons by her mother, he had been left
severely alone. No one had entered the library and the folding doors,
which had been in use for the inquest in place of the portières, and
closed again by Mrs. Meredith after Anne’s departure, had prevented
his hearing anything transpiring in the hall. The clock on the mantel
had ticked off the minutes with maddening regularity. At the stroke
of ten he laid on the smoking table, by his elbow, a box of matches,
which he had been twiddling between his fingers, and picked up his
cane. The opening of one of the library doors caused him to face in
its direction.
“Excuse me, sir,” apologized Herman as he advanced further into
the room, “I did not know you were still here, sir. I was thinking of
closing up the house for the night.”
“I won’t detain you,” replied Curtis quickly. “I am on my way to
bed now. Has every one retired?”
“Yes, sir.” Herman busied himself closing one of the long French
windows opening on the veranda and bolting the other four. “Mr.
Armstrong has just come back.”
Curtis paused on his way to the door. “Mr. Armstrong,” he
repeated, inquiringly. “Mr. Gerald Armstrong?”
 “Yes, sir.” Herman dusted off his hands with a deprecatory
gesture. “He told me, sir, that he missed his train, so he came back,
sir, to spend the night.”
“Oh!” Curtis’ ejaculation covered doubt. He caught and wondered
at the badly suppressed excitement in the butler’s usually
unemotional voice. “Where is Mr. Armstrong?”
“He went straight to his old room, sir; he hadn’t taken away his
things.” Herman switched off two of the tall standing lamps, leaving
the room in semidarkness. “Said I need not disturb Mrs. Meredith to
tell her of his arrival. Is there anything I can do for you, sir?”
“No, thanks.” Curtis reached the doorway and turned around.
“Good night, Herman.”
“Good night, sir.” Herman watched the tall, erect figure pass into
the hall, a glint of admiration in his eyes. “He beats all,” he muttered
under his breath, then devoted his attention to closing the house.
As Curtis reached the staircase a thought struck him and he
hesitated. Why not get Herman to refill his cigarette case from the
stock which John Meredith had kept for his guests? He swung
around and had partially retraced his steps when he paused
abruptly. He had caught the sound of heavy breathing on his right,
then light, receding footsteps.
“Herman?” His low call met with no response, and after a
moment’s wait he returned to the staircase and slowly mounted it,
his cane swinging at a convenient angle in his right hand. It was
leaded and made an excellent club in an emergency.
Keeping his left hand on the banisters, he circled the corner of the
staircase, recalling McLane’s clear description of the way to his
bedroom. He had just made the turn into his corridor when a hail
from Sam Hollister stopped him.
“Hello, Curtis!” Hollister kept his usually hearty voice at a low
pitch. “I am glad you haven’t gone to bed. I want a word with you.”
“You can have more than one if you wish,” responded Curtis. “I
am in no hurry.”
 “Good! Suppose we go to John’s old bedroom. This way.” He
slipped his arm inside Curtis’ and suited his step to his as they went
down the winding corridor. “I was on my way to look you up.”
“Yes?” queried Curtis, as his companion ushered him into the
bedroom, switched on the light and then closed the hall door. “What
can I do?”
An answer came from an unexpected quarter. “Go to H—l!”
shouted Ruffles, awakened from slumber by the brilliant electric
light. The parrot hopped about on his perch and flapped his wings in
Hollister’s face as the latter approached.
“I’ll wring that bird’s neck some day,” he grumbled. “How John
stood his infernal talking is one of the mysteries of this place.”
Curtis snapped his fingers and hummed a popular tune. Ruffles’
plumage assumed its normal sleek appearance and his anger
subsided. He gently nipped Curtis’ extended finger, then with one
sleepy eye cocked at Hollister, descended from the top of his perch
to a lower crossbar and prepared to enjoy his interrupted nap.
“Hum! You seem to have the same knack of pleasing Ruffles as
John,” commented Hollister, eyeing the parrot with disfavor. “Come
over this way, Curtis.” He pushed a chair aside and Curtis followed
him across the bedroom. He judged they were near an open window
from the cooler air which blew upon them. “I’ll shut this in just a
minute—”
“No, please don’t,” broke in Curtis. “The room is a trifle close and
the fresh air feels good.”
“Well, if it’s not too much draught.” Hollister looked somewhat
dubious; he was not a cold-air enthusiast. “Take this seat by the
secretary, I’ll sit here.”
A second later Curtis heard the jingle of keys knocking against
wood. Hollister caught his inquiring expression.
“I’m going through John’s desk,” he explained. “Inspector Mitchell
and Coroner Penfield said they ransacked it thoroughly, without
results, however.”
 “And what do you expect to find?” asked Curtis.
“The documents John signed last night,” promptly. “Or if not,
some clue to their present whereabouts. We could find no trace of
them in the bureau or highboy. This,” laying his hand on the
secretary, “is the only available place for John to place the papers.
He certainly did not leave them lying around the room.”
“Perhaps he gave them to some one,” suggested Curtis, as
Hollister inserted a key in the top drawer of the secretary.
Hollister twisted and turned the key before he could get the
drawer unlocked. “If John did that, wouldn’t that person come
forward now and turn them over to me or to the police?” he asked.
“Provided that person has heard of Meredith’s death,”
supplemented Curtis.
Hollister turned his head and stared at him. “Not know of his
death!” he ejaculated in astonishment. “If John did give them to any
one, that person is living here now. You will recall that no one was
admitted to this house after Gerald Armstrong’s departure.”
Curtis tapped his cane thoughtfully. “I do not recollect that the
coroner asked if any caller was admitted to the house after Meredith
retired to his room,” he said.
“Maybe he didn’t,” retorted Hollister. “But you know that no one
called here, for you were down in the library later than anybody else,
and the library is near the front door—”
“And I am blind.”
Hollister looked taken aback. “I forgot,” he mumbled. “But you
have remarkable hearing—”
“The heavy portières were drawn and I sat in the far end of the
library, near the fireplace,” Curtis pointed out. “Also, I was absorbed
in my thoughts. I cannot swear that no one was admitted last night.”
Hollister took out, examined, and replaced the contents of the
drawer before answering.
“It hadn’t occurred to me that some one—some outsider—might
have had access to John last night after we left him,” he admitted
slowly. “Frankly, I have been haunted by one idea—that the papers
were stolen—”
“By whom?” Curtis’ quiet voice gave no hint of the anxiety
consuming him as he waited for Hollister’s reply.
Hollister carefully sorted a bundle of papers and put them back in
one of the pigeonholes. “By the person who benefited through the
disappearance of the documents,” he said, and Curtis frowned at the
indirect answer.
“And who is that?” he asked.
Hollister eyed him keenly. “You know as well as I,” he exclaimed
roughly. “None other than John’s niece—Anne Meredith.”
Curtis bent the cane in his strong grasp, then let it spring back.
“Miss Meredith asked to have you retained as her lawyer,” he said.
“As her representative you should be the last person to point
suspicion toward her.”
“As her lawyer I am trying to divert suspicion from her by finding
those cursed documents,” snapped Hollister, his quick temper rising.
“And look here, Curtis,” swinging toward the blind surgeon in his
excitement, “it is going to be d—mned serious for her if we don’t
find them. Don’t forget that John was murdered.”
“By heaven! Do you mean to insinuate—”
Curtis was on his feet, his hand clenched about the other’s arm.
“No, no. Let go, you fool!” Hollister strove to free himself. “I
haven’t the faintest idea that she murdered her uncle, but,” as Curtis
released his grip on his arm, “but I do believe that she took those
papers.”
Curtis mastered his temper with difficulty. “Your reasons for
thinking Miss Meredith a thief?” he demanded.
Hollister’s appraising glance at his companion lasted fully a
minute.
“Well,” he said finally, “through the disappearance of the codicil
and the prenuptial agreement, Anne inherits a large fortune without
having to go through a marriage ceremony with you.”
 “And is that your only reason for thinking she took the
documents?” persisted Curtis.
“Isn’t that enough?” replied Hollister, insolently. “She wanted to
dodge being married to you. That, depend on it, made her a thief.”
“Indeed?” Curtis laid his cane across his knees and bent a little
forward. “Then how do you account for the fact that she is still
willing to marry me?”
Hollister dropped the loose papers he had at that moment
removed from a smaller drawer of the secretary.
“Do you mean to say that you two are going on with Meredith’s
plan?” he stammered. “You are going through with the marriage
farce?”
Curtis bowed affirmatively. “We are,” he said. “Anne and I have
just reached that decision.”
“I’ll be everlastingly blessed!” Hollister sat back and contemplated
his companion in astonishment. It was some seconds before he
spoke. “Anne is a damned sight cleverer than I thought!”
“I don’t get your meaning?”
“You don’t, heh? Well,” Hollister pulled himself up short, “let’s see
what we can find in this desk.” He stooped over and picked up the
papers which he had dropped some moments before. “Receipted
bills, household accounts,” running his eyes down them. “Stop a
minute, what’s this?” He unfolded as he spoke a legal-size sheet.
“Evidently part of an inventory, furniture and so on. Here’s a notation
in one corner, written crisscross, in John’s hand: ‘Contents of safe
deposit belongs to’—that’s all,” looking up blankly at Curtis.
“Well, what about it?” asked Curtis, with growing impatience.
“Oh, nothing.” Hollister refolded the paper, gathered the others in
a neat bundle and replaced them in the drawer, but the legal-size
sheet with its inventory of “furniture and so on,” he slipped inside his
coat pocket.
“Put it back,” advised Curtis sternly. Hollister’s mouth dropped
open and his hand fell to his side.
 “How?” he began, then turned fiercely on Curtis. “Damn it, you
can see!”
“No.” Curtis smiled. “You simply forgot that that grade of paper
rustles badly. It required no particular art of divination to detect you,
but don’t try to fool me again, Hollister.”
The lawyer colored hotly, bit his lip, hesitated, then took out the
paper and put it with the others in the drawer.
“I kept it out on impulse,” he said apologetically. “I don’t know
why, unless it was that John’s handwriting in that notation seemed a
bit shaky.”
“Was there room to complete his sentence?” Hollister took the
paper from the drawer again and extended it toward Curtis. “Feel
here,” he said, and guided Curtis’ fingers over the lower right-hand
corner. “What do you find?”
“That the corner has been cut off diagonally,” replied Curtis. He
ran his hand over the sheet. “The other corners are untouched.”
“Just so.” Hollister crossed his short legs and assumed a more
comfortable attitude. “Well, the notation is just above the corner and
runs from edge to edge of the paper. It reads: ‘Contents of safe
deposit box belongs to’—the name must have been written just
beneath it.”
“And cut off.” Curtis handed back the paper. “Put it away, Hollister.
The question now is, did Meredith cut off the corner or did some one
else? And if so, with what object?”
“And what has the contents of the safe deposit box to do with
John’s murder and the disappearance of the codicil and the
prenuptial agreement?” demanded Hollister, his excitement
mounting.
“The answer to that will be found when his safe deposit box is
opened,” replied Curtis dryly. “Does Coroner Penfield know of this
safe deposit box?”
“I told him that John had a box at the Metropolis Bank,” answered
Hollister. “We have taken steps to have it opened in the presence of
the Registrar of Wills and the bank officials to-morrow morning.”
“Good!” Curtis leaned forward and placed the inventory sheet in
the open drawer, then closed it. “Go ahead, Hollister, and look
through the desk.”
“There is only one drawer more that I haven’t examined.” The
lawyer opened it as he spoke and went over its contents with care.
“Pshaw! nothing but invitations, souvenirs, and menus.” He closed
the drawer with a slam. “Our hunt is a failure, Curtis.”
Curtis pushed back his chair. “It would seem so,” he admitted, “as
far as locating the missing papers is concerned. Tell me, Hollister,” as
his companion rose, “what was the relationship between John
Meredith and Gerald Armstrong?”
“Why, none,” responded Hollister. “Armstrong is a man of about
thirty, I should imagine. He is a partner of Colonel Hull’s and that
threw him more or less in contact with John in a business way, as
Hull’s firm transacted some financial deals for John at one time.”
“Is Armstrong particularly attractive?”
“I believe he is quite a favorite with women.” Hollister’s tone
lacked enthusiasm. He paused by the electric light switch,
preparatory to turning it off, when Curtis, who followed him more
slowly across the bedroom, should have reached the hall door. “John
liked him well enough. They always appeared friendly, and he was a
frequent visitor here. I can’t understand why Armstrong left so
suddenly last night, or why he hasn’t been back.”
“Armstrong returned just before I came upstairs.”
“He did?” Hollister stared at Curtis in silence for a second, then
spoke with more than usual rapidity. “Have you talked with him?”
“No. Herman told me of his arrival and that he had gone at once
to his bedroom.” Curtis paused by the open door and, unseen by
Hollister, who had partly turned his back to switch off the lights in
the bedroom, laid his hand on the outside door knob. From it still
dangled the piece of string which the night before had led him to
believe that he was entering his bedroom. “Coming, Hollister?”
 The lawyer closed the door tightly behind him. “I’ll walk with you
to your bedroom,” he half whispered. “It is later than I thought.”
Their footsteps made no noise on the heavy carpet and they
traversed the corridor in silence. At the entrance to Curtis’ bedroom
Hollister bade him a low voiced “good night.”
“Just a second.” Curtis stopped him as he was about to turn back.
“Can I borrow a cigarette?”
“Certainly, take these,” and the lawyer thrust a package into his
hand. “No, I don’t want any to-night,” and not waiting to hear Curtis’
words of thanks, Hollister hurried away.
The package had been thrust into his hand upside down, and to
Curtis’ dismay the cigarettes scattered on the floor before he could
catch them. Stooping down he groped around and after some
difficulty located the majority of them. He was about to rise when he
touched a string partly tucked out of sight under the edge of the
strip of carpet which ran the length of all the corridors.
Getting to his feet, Curtis closed his door, then stooped over. The
bit of string lay in the corridor directly under the door knob.
Curtis carried the string into his bedroom, closed the door, and
making his way to a chair, sat down. First laying aside his cane, he
lighted a cigarette, then held up the string and felt it carefully. He
judged it to be about six inches in length, of ordinary twine, and one
end was tied in a loop which had been neatly cut. Curtis held the
two ends of the loop together. Its size proved that it could have
been tied over his door knob.
Curtis smoked for many minutes without moving, the twine held
suspended in his left hand, and his mind busy with the enigma of
the two strings. Why had Fernando denied tying a string to his door
knob, so that he, Curtis, might identify his bedroom? Why had the
string been cut off, and why, above all, had a string been tied to
John Meredith’s door knob? An hour later Curtis undressed and went
to bed with the enigma still unsolved.
 CHAPTER X

THE SOLITARY INITIAL

Gretchen, the chambermaid, craned her neck over the banisters in

her endeavor to find out what was going on in the large square
reception hall on the floor below. Her limited knowledge of English
prevented her understanding much of what she overheard. The
voices grew more indistinct as the speakers moved away, and finally
ceased entirely. Gretchen straightened up and rubbed her stiff
muscles, then with a backward glance down the corridor toward Mrs.
Meredith’s boudoir door, she turned to her right and ran into
Susanne.
“Oh, excuse!” she exclaimed in confusion, her pretty color
mounting.
Susanne picked up the lingerie which Gretchen’s unexpected
collision had knocked from her hand and smiled kindly.
“Mon Dieu. You are in a hurry,” she commented. “But, petite, why
so white?” as Gretchen’s color receded as rapidly as it had come.
“I”—Gretchen caught her breath sharply—“it is this house; it make
me nervous.” Tears hung on her eyelashes and she brushed them
away. She edged closer to the French maid, who was eyeing her in
real concern. “Did you go with madame to the funeral?”
“But, yes.” Susanne’s kindly expression altered to one of deep
seriousness. “The services were of the most simple at the chapel,
but at the grave were many strangers and they crowded about until
some one in authority ordered them back. Mademoiselle Anne was
greatly upset and Madame Meredith very angry.”
“Have they returned?” questioned Gretchen timidly.
“Oui. Here comes madame now,” as Mrs. Meredith’s voice was
heard on the staircase. With an alarmed look behind her, Gretchen
darted past Susanne and down the corridor toward the back stairs.
The French maid regarded the flying figure thoughtfully for a
second, then advanced in time to meet Mrs. Meredith at the head of
the circular staircase.
“Shall I pack madame’s trunks?” she asked as the older woman
paused to take breath after her rapid climb upstairs.
“No; Mr. Hollister wishes us to remain here for several days
longer,” replied Mrs. Meredith. “Have you seen Miss Anne?”
“Oui, madame; she has gone to her room.” Susanne followed her
mistress down the corridor. “Doctor McLane just telephoned,
madame, that he call soon to see Mademoiselle Anne.”
“Very well, let me know first,” with emphasis, “when he comes.
Wait for me in my bedroom, Susanne,” and Mrs. Meredith crossed
the boudoir. Not stopping to knock on the panel of the closed door,
she opened it, and stepped inside her daughter’s room. Anne looked
up from the couch where she had thrown herself twenty minutes
before, and at sight of her mother, half rose.
“Don’t get up.” Mrs. Meredith drew a chair over to the couch and
seated herself. At her air of conscious rectitude Anne’s heart sank.
“There is something I wish to discuss with you.” Unconsciously Anne
braced herself; her mother’s “something” was sure to be
disagreeable—it generally was. “I noticed, Anne, that during the
funeral services you sat in the same pew with Doctor Curtis.”
“Yes, mother, I did.” Anne judged she was expected to answer as
Mrs. Meredith came to a full pause.
“And you took his arm and walked with him afterward from the
chapel to the grave?”
“I did.” She gazed full at her mother. “He is blind, you know.”
 “So that was it—philanthropy.” Mrs. Meredith nodded her head,
well satisfied. “But, my child, don’t let your kind heart run away with
our discretion. It is no longer necessary to cultivate Doctor Curtis’
acquaintance.”
“I beg your pardon, mother.” Anne’s heart was beating a bit more
rapidly. “I do not agree with you.”
Mrs. Meredith sat back in her chair. “When you take that tone,
Anne, I know you are going to be obstinate. But you must listen to
me. The so-called ‘engagement’ between you and Doctor Curtis is at
an end.”
“On what grounds?” meeting her mother’s eyes. “Expediency?”
“Anne, how dare you?”
Anne straightened her slender figure and threw back her head.
“On the contrary, mother,” she said clearly, “Doctor Curtis and I will
carry out Uncle John’s plans to the letter.”
Mrs. Meredith gazed at her, thunderstruck. “You mean—”
“That our marriage will take place before this week is out.”
Mrs. Meredith, livid with wrath, sat for some moments absolutely
silent. When she finally spoke, both voice and manner were more
conciliatory.
“Heroics are all very well in their place,” she began, “but before
this rash marriage is consummated, there are many things to
consider. First, Doctor Curtis is blind. He has no future,” she paused,
“but he has a past—”
“Explain your hints, mother,” as Mrs. Meredith paused again.
“Has he spoken to you of his past career?”
“No.” Anne’s white cheeks turned crimson. “We have never had a
lengthy conversation.”
“It is just as well,” dryly. “I have started an investigation—”
Anne was on her feet, her usually calm, cold demeanor
transformed into passionate fervor. “I warn you, mother, to stop any
so-called investigations. Is your record, and mine, so clean in this
plan for a hurried, wild marriage that we can afford to blacken the
man, who under hard pressure of blindness and destitution
consented to it?”
“Anne!”
“Stop, mother; I will be heard,” as Mrs. Meredith raised her hand
with an imperative gesture. “Doctor Curtis afforded us the means to
gratify that mysterious mandate which Uncle John insisted upon by
agreeing to marry me, and by that marriage, in name only, I will
inherit a large fortune.”
“Your uncle’s death alters that—”
“Does it?” For the first time Anne did not meet her mother’s eyes.
“Doctor Curtis has proved himself a gentleman and a man of honor
in his treatment of me. Yesterday, when I was heckled by Coroner
Penfield, he came to my assistance. I,” raising her head proudly, “I
will not be a party to any act, overt or concealed, which endeavors
to pry into his past.”
The door banged shut as Anne, springing to her feet, fled through
it. Pressing her hands against her hot cheeks, she leaned panting
against the wall of the boudoir to recover her self-possession before
going to Lucille’s bedroom.
Downstairs in the library Sam Hollister rubbed his bald head with a
large silk handkerchief and gratefully accepted Herman’s suggestion
of a cocktail from what had once been John Meredith’s private stock.
“Bring three,” he added. “I am sure Mr. Armstrong will join me,
and Doctor Curtis will be here presently.” As the butler disappeared,
he turned to Gerald Armstrong. “A cocktail,” he remarked dryly, “may
make you a more agreeable companion.” Armstrong transferred his
gaze from his carefully creased trousers to Hollister’s flushed
countenance.
“Why so heated?” he asked. “Sit down and take things calmly.”
The look that the lawyer cast at his younger companion was
anything but complimentary. “Calmly?” he fumed. “Where is that
ass, Hull?”
 “Do you mean Colonel Julian Hull?” Armstrong made no attempt
to conceal his amusement. “My revered senior partner is, I believe,”
glancing at his wrist watch, “in our office watching the stock
market.”
“And you ought to be with him,” with equal vehemence. “Why are
you hanging around this house?”
“Isn’t that my business?” Armstrong’s sallow cheeks had turned a
deep red, but otherwise he displayed no anger. His voice had not
lost its teasing quality, which to many people was an annoying
characteristic.
“It may be the coroner’s business if you are not careful,” exploded
Hollister, losing his little store of patience, which had been sorely
tried that morning. “What put it into your head not to appear at the
inquest yesterday afternoon?”
“My dear Hollister,” Armstrong smiled tolerantly, “I explained in my
note to Coroner Penfield, which I understand he did me the honor to
read at the inquest, that not having been in the house at the hour
John Meredith died, my testimony would add nothing to the
investigation.”
His voice carried to the farther end of the library and David Curtis
listened attentively as he skillfully avoided the furniture in his slow
progress toward the two men. Absorbed in watching each other,
neither man heard his approach. Curtis paused almost at Hollister’s
back and gently struck his cane against the side of a mahogany card
table. Armstrong swept a startled glance behind him and then
resumed his nonchalant pose, while Hollister stepped to one side
and laid his hand on a chair back.
“Hello, Curtis!” Hollister pushed the chair he had grasped toward
the blind surgeon. “Sit down and be comfortable. Here comes
Herman with the cocktails.”
“Thanks, but I won’t have any,” Curtis said, as the butler stopped
before him, silver tray in hand. “Cocktails and brain work don’t go
together successfully.”
 “And what does your brain work comprise?” asked Armstrong,
with a snicker of amusement as he took one of the frosted glasses.
He drained his before Herman had time to serve the lawyer. “If you
don’t wish the one Doctor Curtis scorned, I’ll take it, Hollister.” He
drank the second cocktail more leisurely, then turned to Curtis. “You
haven’t answered my question, doctor.”
“Ever hear of the fourth dimension, Armstrong?” Curtis smiled, as
he addressed the young stockbroker; it deepened at the latter’s
sulky nod. “Well, a problem of that kind provides very good mental
relaxation—”
“For a blind man,” interjected Armstrong, contemptuously.
“Just so,” agreed Curtis, his manner unruffled. He turned to their
silent companion. “Why so fidgety, Hollister? You have snapped your
watch cover shut half a dozen times since I have been sitting here.”
Hollister replaced his hunting-case watch in his pocket.
“Mrs. Meredith is late,” he explained. “We have to be at the
Metropolis Bank in twenty minutes.”
Armstrong leaned forward, a touch of eagerness in his manner.
“So you are going to open John Meredith’s safe deposit box to-
day,” he commented. “I understand the bank officials had called it
off until later in the week.”
“I don’t know who your informant could have been,” replied the
lawyer dryly, “but it has not been postponed, except as to hour, to
oblige Coroner Penfield. Ah, here is Mrs. Meredith,” as the widow
appeared in the doorway.
“Don’t let me disturb you, Gerald,” she exclaimed, as Armstrong
went with Hollister to the door. “Oh, Doctor Curtis, I did not at first
see you,” catching sight of the blind surgeon over Hollister’s
shoulder. She turned to the lawyer. “I am sorry to have kept you
waiting, Sam; but Anne detained me. How long do you suppose we
will be at the bank?”
“About an hour, perhaps two, but not longer than that,” Hollister
added, catching her expression of dismay.
 “In case we are delayed in returning,” Mrs. Meredith addressed
Curtis directly, “I have told Herman to serve luncheon and not to
wait for us. In our absence, doctor, I trust that you will act as host.”
“Thank you, Mrs. Meredith,” replied Curtis, bowing deferentially.
He could not see the sudden look of aversion which Gerald
Armstrong cast in his direction, but he was aware intuitively that
Mrs. Meredith’s formal courtesy cloaked the animosity which he fully
realized from almost their first meeting was only slumbering, ready
to burst forth at any moment. That she had not taken kindly to his
inclusion in the house party had been incautiously told him by Lucille
Hull; and he judged that only dire necessity had later induced Mrs.
Meredith to agree to her brother-in-law’s plan for his marriage to her
daughter.
Herman’s approach broke up the little group. “Damason is at the
door, madam, with the car,” he announced, and with a bow to Curtis
Mrs. Meredith moved away, Hollister in her wake. Armstrong was
about to follow them when Herman addressed him.
“Inspector Mitchell has just telephoned to ask if you were here,
sir,” he said. “He is waiting to speak to you.”
Armstrong smothered an oath. “Tell him to go to—Guinea!” he
directed. “No, wait,” as Herman bowed and moved a few steps away.
“I’ll talk to the beggar,” and he hurried back into the library, and over
to the branch telephone standing on a small table in a corner, which
had been devoted exclusively to John Meredith’s use.
Armstrong’s conversation over the telephone with Inspector
Mitchell appeared to be a strictly one-sided affair, or so Curtis judged
from the few monosyllabic remarks from the stockbroker. When he
hung up the receiver a few minutes later he was scowling.
“Persistent devils, these detectives,” he said, walking over to the
smoking stand and striking a match which he applied to an
expensive cigar. “Mitchell insists that I wait until he gets here.”
“Does his request put you to inconvenience?” asked Curtis politely.
Armstrong shrugged his shoulders, but whatever answer he would
have made was forgotten on catching sight, through one of the
windows, of Lucille and Anne walking across the lawn toward the
lodge. Without a word of explanation to Curtis, he opened the
French window and hurried after the two girls.
Curtis made his way over to the window and stood in it facing the
lawn. He was not aware that his tall figure in its well-fitting suit of
gray clothes was silhouetted against the dark background of the
library, or that, at Armstrong’s hail, Anne and her cousin had swung
around. Anne’s gaze traveled past Armstrong’s advancing figure and
rested on Curtis. She instinctively raised her hand to wave a friendly
greeting, then dropped it. For an instant she had forgotten that
Curtis was blind. There was a catch in her throat as she spoke to
Armstrong and her face was unsmiling as she walked with him and
Lucille to the lodge.
It was fully ten minutes before Curtis left the window and went
slowly upstairs to his bedroom. Pausing by his bed, he laid his cane
across it. In doing so his hand touched some clothing. Lifting it up
he found it was a suit of pajamas. Curtis bent down and passed his
hand rapidly over the bed; it was, as he thought, made up. Why
then were his pajamas laid out on the bed at noon? Had Gretchen,
the chambermaid, forgotten to put them away or was it carelessness
on the part of Fernando, his Filipino valet?
Somewhat perplexed, Curtis again picked up his pajamas. As he
ran his fingers over the jacket he drew out a handkerchief from the
pocket. Holding it close to his nose he detected the odor of
chloroform. Only a faint, very faint, trace of the chloroform
remained, but it was sufficient to identify the handkerchief as the
one thrown toward him by the unknown woman in John Meredith’s
bedroom on the night of Meredith’s murder.
Curtis sat down in the nearest chair and spread the pajamas
across his knee. In the rapid march of events he had forgotten the
handkerchief which he had inadvertently stuffed into the pocket of
his pajamas on going to his room to rest after the discovery of
Meredith’s body.
 He judged the handkerchief to be of the finest linen, of dainty
size. Deftly his fingers traveled around its edges. Was there no mark
by which he might establish the identity of the mysterious woman in
Meredith’s bedroom? His long, sensitive fingers stopped at one
corner. Slowly they traced out the solitary initial—the capital letter
“A.”
 CHAPTER XI

THE HAND ON THE COUNTERPANE

A low tap at his bedroom door aroused Curtis. Rising in some

haste he went over to his bureau, took out his despatch box, and,
opening it, securely locked the handkerchief inside it. Not until the
box was again in the drawer did he turn toward the door.
“Come in!” he called as the knock was repeated with more
insistence. Doctor Leonard McLane stepped briskly inside and closed
the door behind him.
“I am glad I found you, Dave,” he said, and, observing Curtis’
pleased smile on recognizing his voice, added: “I called to see Anne
Meredith, but she had gone out motoring with Lucille Hull and
Gerald Armstrong. Herman told me that you were in, so I came
upstairs.”
Curtis sighed with relief. “I am very glad that you are here,
Leonard,” he exclaimed. “Frankly, I was just thinking of telephoning
to you to come over at once.”
“Indeed?” McLane drew up a comfortable rocker and seated
himself near the blind surgeon. “What do you wish to see me about,
and why are you caressing a pair of pajamas?”
As he spoke Curtis had picked up the pajamas from the chair
where he had dropped them upon hearing McLane’s knock on his
door.
“I’ll explain all in good time,” he answered, seating himself.
“Please treat our conversation as confidential, Leonard.”
 McLane nodded his head thoughtfully. “I presume it’s about John
Meredith’s murder and”—he hesitated—“Anne.”
“Why do you connect the two?” quickly.
“It is what every one is doing,” said McLane. He noticed the
harassed lines in Curtis’ face and his expression grew more serious.
“Coroner Penfield told me what transpired at the inquest and that
you insisted that Anne be represented by a lawyer. How,” he glanced
keenly at his companion, “how did you happen to pick on Sam
Hollister?”
“Anne asked for him,” replied Curtis. “Isn’t he a good lawyer?”
“W-why, yes; so I understand.” McLane’s tone did not convey
conviction. “But he is not a criminal lawyer.”
Curtis hitched his chair closer to McLane. “You think it will come to
that?” he asked, with unconcealed anxiety.
McLane nodded his head somberly. “It appears to me that Anne
knows more than she has told,” he said. “Why she is withholding
information which may aid the police in detecting her uncle’s
murderer is one of the mysteries of the case.”
“But there is no criminal action in that,” protested Curtis.
“Unless it comprises being an accessory after the act,” McLane
pointed out. He paused a moment before asking, “What are the
known facts connecting Anne with the murder?”
Curtis sat back in his chair and checked off each point as he
spoke. “First, Herman, the butler, testified that he overheard John
Meredith quarreling with a ‘female’ in his bedroom that night. He
took her to be Anne because he thought he recognized her dress.
Secondly, Gretchen, the chambermaid, said that she overheard a
conversation between a man and a woman after midnight under her
window. The woman said, ‘I will do it to-night,’ and the man replied,
‘Don’t lose your nerve.’”
“Well, did Gretchen identify the woman?” asked McLane as Curtis
paused.
 “Indirectly, yes. She declined, as she put it, ‘to tell on her young
mees.’” Curtis hesitated. “Her statement satisfied the coroner and
she was excused.”
“I see!” McLane stroked his chin reflectively. “Well, what next?”
“I overheard Mrs. Meredith speak to Anne in the hall just after I
found Meredith’s body.” Curtis spoke with growing reluctance, and
McLane nodded his head in silent understanding. “Mrs. Meredith said
nothing to connect Anne with the crime, but it did prove that Anne
was up and about at the time of her uncle’s murder.”
“Quite so, it did,” agreed McLane. He lowered his voice. “Did
anything come up at the inquest about the parrot and its cry: ‘Anne
—I’ve caught you—you devil?’”
“No.”
McLane sat back and frowned. “Why not, I wonder?” he muttered.
“The inquest is not over,” Curtis pointed out. “Only adjourned until
Thursday.”
“And this is Tuesday morning—”
“Which leaves us very little time to solve the mystery of Meredith’s
death.” Curtis sighed, then bent forward and laid his hand on
McLane’s knee. “Can I depend upon your help, Leonard?”
“Absolutely.”
“Good!” Curtis’ face lighted with his charming smile. “We must
work to clear Anne. She must not be dragged any further into the
limelight.”
“If it only stops at the limelight!” The exclamation escaped McLane
involuntarily. “I am afraid, Dave, that Coroner Penfield is holding
back something more than the episode of the parrot to spring at the
next hearing of the inquest.”
“It may be,” admitted Curtis. “Penfield stopped his direct
examination after producing the hair which he and Inspector Mitchell
found wound around the button on the jacket of Meredith’s pajamas.
The hair matched Anne’s in color and texture.”
 “And Penfield claimed that it was caught around the button when
Anne pressed her ear over Meredith’s heart to see if it was still
beating,” broke in McLane. “It was a clever deduction on his part.”
“Quite so, and one warranted by facts—as far as he knew them,”
answered Curtis. “Is the hall door closed, Leonard? Are we alone?”
McLane glanced toward the door and then about the room.
“The door is shut,” he said. Rising, he walked over to it, pulled it
open and glanced up and down the empty hall, then closed the door
and turned the key in the lock. “We are entirely alone, Dave. Go
ahead and say what you wish.”
Curtis waited until his companion had resumed his seat.
“After I had notified Sam Hollister of Meredith’s death, I went back
with him to the body,” he began. “Hollister left me to telephone to
Coroner Penfield. While waiting for him to return, I ran my hands
over Meredith’s body and found some hair, evidently from a woman’s
head, caught around that jacket button.” He paused. “I may also
state that when I first found Meredith he was lying partly on his right
side, face pressed against the carpet and his arms outflung.”
“So I read in your printed testimony,” interjected McLane.
“But when I examined the body for the second time, it was lying
on its back,” finished Curtis.
“It was?” McLane shot a questioning glance at his blind
companion. “Why didn’t you mention it at the inquest?”
“I was not questioned on that point,” calmly. “If I am recalled at
the next hearing I will speak of it. In the meantime—”
“Yes?” as Curtis paused.
“I want your advice, and, if need be, your aid.”
“Sure, go ahead.” McLane was listening with deep attention and
with increasing impatience at his friend’s deliberation of speech.
“I unwound several of the hairs,” went on Curtis, “and put them in
my wallet. Later that morning, that is yesterday, I showed them to
Fernando and asked him their color. He said the hairs were white.”
“White!” echoed McLane in astonishment.
 “Fernando said that they were white,” repeated Curtis. “I had to
depend upon his eyesight.”
“But,” McLane took out his handkerchief and dabbed his forehead,
“the hair Penfield found about the button was chestnut in color. I’ve
seen it and it certainly came from Anne’s head.”
“Possibly Fernando lied when he told me the hair was white.”
“Sure, he might have; with the object of shielding Anne. The
servants are devoted to her,” McLane added. “Let me see the hair
and I can settle the question.”
“Unfortunately the hair has disappeared out of my wallet.”
“Good Lord!” McLane sat back and regarded Curtis in startled
surprise.
“I discovered it was missing during the inquest at the time it was
stated that the hair Penfield found was chestnut in color,” went on
Curtis. “Having nothing to prove my statement, I kept silent.”
“I see!” McLane gnawed at his upper lip. A second or more passed
before he broke the silence. “There isn’t a white-haired woman in
the household,” he said.
“Then Fernando lied,” Curtis’ lips compressed into a hard line, “and
not for the first time. Listen attentively, Leonard.” The injunction was
hardly needed, but Curtis could not see his companion’s absorbed
regard as he sat back watching him. “When dressing for dinner on
Sunday evening I told Fernando to tie a string on the outside knob
of my door so that when I came upstairs, if I was alone, I could
identify my bedroom without difficulty.”
“Did he do so?”
“No. Fernando claims that I never ordered him to tie a string on
the door knob.” Curtis spoke more slowly than usual. “But after
discovering Meredith lying dead in the hall, I went in search of my
room and, finding a string hanging from a knob of a closed door,
entered that bedroom, supposing it to be mine.”
“Whose was it?”
“John Meredith’s.”
 McLane sat back and again rubbed his forehead with his
handkerchief.
“I’m blessed if I see—” he exclaimed.
“Unhappily I don’t see—at any time.” Curtis covered his sigh with
a slight cough. “This is the point, Leonard; a string was tied to John
Meredith’s door knob and is still hanging there. A string was also
hung on my door knob Sunday evening and cut off before I came
upstairs.”
“What?”
Instead of replying Curtis rose and went over to his bureau.
Taking his despatch box from the drawer he made his way to the
bed and, turning the key in the lock, threw back the lid.
“This piece of string,” he said, holding it up, “has one end tied in a
loop, which has been cut.” He handed the string to Leonard. “I found
the string lying in front of my door, partly hidden under the hall
carpet.”
McLane took the string and eyed it attentively. “Just a moment,”
he exclaimed. “I’ll be back.” He stopped at the hall door, unlocked it
and sped up the hall. During his absence Curtis stood by the bed,
head bent in a listening attitude. Barely three minutes elapsed
before McLane was beside him again.
“I have compared the string with that still hanging from Meredith’s
door,” he said, in explanation. He placed the string in Curtis’ hand.
“It is the same color and weight, and was evidently cut from the
same ball of twine.”
“And Fernando denies that I ever requested him to tie a string to
my door,” mused Curtis, as he put the string back in his despatch
box.
“Could he have tied the string on your door, then cut it off, and
tied one on Meredith’s door as a practical joke?” asked McLane. “And
after the events of Sunday night be afraid to confess?”
“That is a plausible theory,” admitted Curtis, somewhat dubiously,
however. “But why pick out John Meredith’s door?”
 “Ask me something easy,” begged McLane. “Did you go in
Meredith’s bedroom, Dave?”
“Yes. I telephoned from there for Sam Hollister.” Curtis paused,
then spoke with added gravity. “While standing before the
instrument trying to recall Hollister’s number, I heard a woman
moving about in the bedroom.”
McLane’s eyes were twice their usual size. “Go on,” he urged.
“Don’t keep me in suspense. Did the woman see you?”
“No. I had not switched on the electric lights,” Curtis explained,
keeping his voice low but distinct. “As she went by me on her way
out of the room, she tossed this handkerchief in my direction.” He
took it out of his despatch box and gave it to McLane. “When I
picked it up I detected the smell of chloroform very plainly.”
McLane turned the handkerchief over several times and the
solitary initial caught his eye.
“A,” he said aloud, and the gravity of his tone was unmistakable.
“Anne?” He laid the handkerchief back in the despatch box. “Lock up
the box, Dave,” he directed. “Have you shown the handkerchief to
Coroner Penfield?”
“No.” Curtis pocketed the key of the despatch box. “I know you
won’t approve, Leonard, but”—and his tone was grim—“I decline to
further involve Anne Meredith in the mystery of her uncle’s murder.”
“I am with you there,” declared McLane. “I wish, however, that
you had spoken to me sooner about the handkerchief.”
“This is the first time I have seen you since we met in Meredith’s
bedroom yesterday,” Curtis pointed out. “But I must confess,
Leonard, that the handkerchief did slip my mind. I had left it in the
pocket of this suit of pajamas, and only recollected the handkerchief
when I found the pajamas lying on this bed about fifteen minutes
before you came in.”
“Lazy habits you have,” commented McLane, speaking more
lightly. “Leaving your pajamas around your room at this time in the
morning.”
 “I did not leave them there,” protested Curtis. “I don’t know who
could have laid them on the bed. It’s made up, is it not?”
McLane turned about and gazed at the bed as Curtis crossed the
room to his bureau, despatch box in hand.
“The bed is made up,” McLane stated slowly. Something caught his
eyes and he stepped close to the bedstead and bent forward. “By
Jove!” he exclaimed. “There is an impression of a hand on the
counterpane—”
“Monsieur le docteur!” McLane straightened up swiftly and
encountered Susanne’s frightened gaze. The French maid was
standing holding the hall door ajar. “Mademoiselle Anne is calling for
you—come quickly!”
 CHAPTER XII

MURDER

David Curtis was not far behind Leonard McLane in reaching the
hall and instinctively swung in the direction the latter was headed.
Anne Meredith turned back from the head of the circular staircase at
their approach.
“Oh, Doctor McLane!” she exclaimed. “I found poor Gretchen
stretched out here in a dead faint. She is coming to, now. Thank
you,” addressing Inspector Mitchell who, seated on the top step,
supported the chambermaid’s head on his broad shoulder. “You were
very kind.”
“Not at all, Miss Meredith.” Mitchell, considerably embarrassed by
his role of nurse, gladly relinquished his place to McLane and
Susanne, who at a sign from Anne helped to support the half-
conscious Dutch girl.
Herman, standing in the square hall at the foot of the circular
staircase, had heard the commotion and, with forethought, instantly
provided himself with a glass of water and a smaller glass containing
whisky.
Armed with these he appeared on the scene just as McLane, with
the assistance of Susanne, had gotten Gretchen stretched out on a
broad settee which stood in a window alcove off the corridor.
Susanne placed a pillow under Gretchen’s head and loosened her
black gown with a deftness which won an approving word from
McLane. It took some persuasion to induce Gretchen to swallow
some of the whisky and she made a wry face as the powerful