TESTIMAGES - a large-scale archive for testing visual devices and basic image processing algorithms

N. Asuni, A. Giachetti
TESTIMAGES: a large-scale archive for testing visual devices and basic image processing algorithms
STAG: Smart Tools & Apps for Graphics - Cagliari, Italy - September 22-23, 2014 1

2
TESTIMAGES
a large-scale archive for testing visual devices and
basic image processing algorithms
● Nicola Asuni
Tecnick.com LTD, UK
● Andrea Giachetti
Dep. Computer Science, University of Verona, Italy
STAG: Smart Tools & Apps for Graphics - Cagliari, Italy - September 22-23, 2014

3
INTRODUCTION TESTIMAGES EXAMPLES DISCUSSION
Motivation
 Necessity of digital images to test, evaluate and optimize basic
image processing software and algorithms, physical displays
with different resolutions and rendering pipelines.
 Necessity of precise geometric and color patterns at different
resolutions.
 Difficulty to find the exact images used by others and be able
to reproduce and compare results.
 Ability to save time by reusing a ready-made dataset.
 Limitations imposed by different image licensing and copyright
laws in different countries.
 Lack of a large common or standard dataset of hi-quality test
images for basic image processing.

4
Another Image Dataset?
 Good image datasets are available but they are not
generally suitable for research due to their unknown source,
poor quality or licensing/copyright issues.
 Numerous research-oriented image archives (e.g.
computervisiononline.com) are only useful or specific
tasks (e.g. PASCAL dataset for semantic annotation, LIVE
for image quality assessment, others for faces, textures,
etc.)
 There are not large datasets for patterns and colors at
multiple standard display resolution and quality assessment
for image rescaling.

5
 Proposed Solution
 Publicly available, large and free collection of digital images
(TESTIMAGES) mainly designed for analysis and quality
assessment of different kinds of displays (i.e. mobile phone
screens, monitors, televisions and digital cinema projectors)
and basic image processing tecnhiques (i.e. resampling,
rendering, perception).

6
Main Features
 Publicly available : http://testimages.tecnick.com
 Free licence : CC BY-NC-SA 4.0
Creative Common licence Attibution-NonCommercial-ShareAlike 4.0 International
 High resolution 16bpp and 8bpp, RGB and Grayscale images.
 Multiple datasets with hundreds of different artificial (computer-generated)
and natural reference images.
 More than 2 million ready-made images, counting the
variations in color, bit-depth and resolution.
 Availability of shifted downscaled images for resampling and
superresolution algorithms.
 Common file naming convention.

7
Archive Organization
 The archive is currently organized in four main sections:
● COLOR
mainly aimed at testing color rendering on different displays
● PATTERNS
mainly aimed at testing the rendering of standard geometrical patterns
● SAMPLING and SAMPLING_PATTERNS
mainly aimed at testing resampling algorithms
 Despite their main categorization, the images can be
easily repurposed for different scenarios.

8
COLOR
 Computer-generated Images mainly aimed to test the
color rendering on different displays and facilitate color
adjustments and calibration.
 7 reference RGB images, including the standard
SMPTE RP 219:2002 calibration image, all available in
16bpp and 8bpp for 114 different standard and common
resolutions up to 16K Digital Cinema (16384x8640).

9

10
 PATTERNS
 Computer-generated images mainly aimed to test the
rendering of standard geometrical patterns by display,
monitors, televisions and projectors, facilitating the
calibration process and the detection of defects.
 38 base patterns with multiple variations, available in
Grayscale and RGB with different color channel
combinations, 16bpp and 8bpp and adapted for 114
standard and common resolutions.
 Helps to optimize or identify specific issues in some
image processing algorithms.
 Some patterns can be used as graphic masks.

11

12
 SAMPLING
 Natural images mainly aimed at testing resampling
algorithms (i.e. interpolation, zooming, enlargment and
superresolution).
 40 RBG 1224x1224 pixels 16bpp HDR reference
images.
 Subresolution and variations with 8bpp and grayscale,
for a total of more 220K images.
 An approach involving multiple camera exposures was
used to acquire 16bpp HDR natural images since there
are not 16bpp capable photographic cameras in the
consumer market.

13

14

15
 SAMPLING PATTERNS
 Aim to test in isolation specific features of resampling
algorithms.
 This is an extension of the SAMPLING archive
containing 424 artificial Grayscale reference images,
generated as described in the PATTERNS section for the
1224x1224 pixels resolution, for a total of more than 1
million image variations.

16
 Evaluation of upscaling algorithms
 The SAMPLING and SAMPLING_PATTERNS datasets
can be used to evaluate upscaling algorithms with
enlarging factors of 2, 3, 4, 5, 6, 8, 10 and 12.
 Subsampled and shifted images are available to
account for the half-pixel shifts introduced by hole-filling
interpolation methods.
original pixels
sub-sampled pixels (C00C00)
shifted interpolated pixels using hole-filling method
(B01R01)

17
Display Calibration
 In the COLOR dataset, the resized variations of the
standard SMPTE RP 219:2002 image can now be used
to calibrate different resolutions. The other images can
be used to calibrate features like color balance, gamma
and saturation.
 The images in the PATTERNS dataset can be used to
calibrate features like contrast, clock, phase, sharpness
and gradient banding.

18
Discussion
 Research and development in Computer Graphics and
image processing may obtain huge benefit from the
availability of public tools for the technical and scientific
community.
 The TESTIMAGES archive is proposed as a ready and
free to use large image dataset with different image types
for different purposes.
 The archive has been already downloaded thousands of
times and used in scientific publications.
http://testimages.tecnick.com

TESTIMAGES - a large-scale archive for testing visual devices and basic image processing algorithms

Recommended

More Related Content

Viewers also liked (8)

Similar to TESTIMAGES - a large-scale archive for testing visual devices and basic image processing algorithms (20)

More from Tecnick.com LTD (6)

Recently uploaded (20)

TESTIMAGES - a large-scale archive for testing visual devices and basic image processing algorithms