0% found this document useful (0 votes)
34 views3 pages

Fused Color Indexed Histogram and Curvelet Based Image Classification and Retrieval System

This document proposes a fused color indexed histogram and curvelet based image classification and retrieval system. It summarizes that content based image retrieval aims to retrieve similar images to a query image but the semantic gap remains. The proposed approach uses a color indexed histogram to extract color features and curvelet transform to extract texture and shape features, which are then classified using support vector machines. The system is evaluated on 1000 images from the SIMPLIcity database and shows improved retrieval accuracy compared to other region based methods.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
34 views3 pages

Fused Color Indexed Histogram and Curvelet Based Image Classification and Retrieval System

This document proposes a fused color indexed histogram and curvelet based image classification and retrieval system. It summarizes that content based image retrieval aims to retrieve similar images to a query image but the semantic gap remains. The proposed approach uses a color indexed histogram to extract color features and curvelet transform to extract texture and shape features, which are then classified using support vector machines. The system is evaluated on 1000 images from the SIMPLIcity database and shows improved retrieval accuracy compared to other region based methods.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

FUSED COLOR INDEXED HISTOGRAM AND CURVELET BASED IMAGE

CLASSIFICATION AND RETRIEVAL SYSTEM

Dr. V. Karpagam
Associate Professor
Department of Information Technology
Sri Ramakrishna Engineering College, Coimbatore
[email protected]
Ph: +919842073310

Abstract

Content Based Image Retrieval (CBIR) finds its application in a wide variety of domains like quality
control, video surveillance, target identification and industrial automation. Myriad of approaches have
been proposed to retrieve similar images to a query image from the image database. But still the semantic
gap remains to be bridged.

Curvelet transform allows an optimal sparse representation of images. Traditional wavelet transforms
represent the point singularities well, but they ignore the geometric properties and therefor e the edge
information present in an image. They suffer from poor directionality properties.

Introduction

CBIR involves retrieving similar images to a query image from the image database. CBIR gains an edge
over text based retrieval approaches as textual annotations of an image can be ambiguous. Myriad of
approaches are found in literature for CBIR but the semantic gap remains to be bridged, hitherto. The
accuracy with which similar images are retrieved from the image database still remains low.

Related Work

Popular Image retrieval systems like SIMPLIcity[ ], BlobWorld[ ], NetRa[] are all region based systems.
Regions based CBIR (RCBIR) systems partition or segment the images into regions and perform a many
to many comparison of the regions. These techniques are computationally expensive. Many of the global
techniques _____________ compromise on the retrieval accuracy.

Proposed Work

The proposed work makes use of a unique color indexed histogram to extract color information from the
image. Curvelet transform is used to extract texture and shape information. The fused features are
classified using Support Vector Machine (SVM). The performance of the system is evaluated using
SIMPLIcity database of 1000 images. The proposed Curvelet Histogram based image retrieval system
with Support Vector Machine based classification (CHSVM) shows significant improvement in retrieval
accuracy in comparison with WHSVM and other region based systems.

Color Indexed Histogram


In the proposed method, the RGB image is converted to an indexed image with low level of color detail.
All the 16777216 possible colors of the RGB space are not perceivable by the human eye. Therefore the
24 bit color image is quantized to a 256 color indexed color image. A common operation that reduces the
size of large 24 bit bitmaps is to convert them to indexed color with an optimized palette, that is, a palette
which best represents the colors available in the bitmap. The color map of only one image of the whole
dataset is stored separately to decompose the remaining images of the data set. A color approximation
method is used to do the color mapping and the images will be almost in their original color. Indexed
color is an economical way of storing color bitmaps without using 3 bytes per pixel. As with 8 bit grey
bitmaps each pixel has a byte associated with it. Only now the value in the byte is no longer a color value
but an index into a table of colors called a palette or color table. Figure 1 shows the color map.

Figure 1. Colormap of Image chosen from database

The color indexed histogram results in an improvement in the representation of the semantic content of
the image, when compared to the ordinary 256 bin color histogram. The discretization into 256 colors for
an ordinary histogram is done based on regular intervals, which may result in different colors being
mapped to the same bin and similar colors being mapped into different bins. The color indexed histogram
proposed in this method derives its color map, from one of the images in the database. Therefore, the
distribution of colors will be natural. Minimum variance quantization is used to quantize the colors into
256. The color index also has the added advantage of being able to reconstruct the image with some loss
in color information, with only one third of the space required to store the original images.

Most natural images exhibit curve singularities.

Curvelet is used to represent images at different scales and different angles. It has strong directional
properties.

The Curvelet Transform includes four stages:

Sub-band decomposition

Dividing the image into resolution layers.

Each layer contains details of different frequencies:

P0 Low-pass filter.

1, 2, Band-pass (high-pass) filters.

The original image can be reconstructed from the sub-bands:

Energy preservation

Smooth partitioning

Renormalization

Ridgelet analysis

Ref:

Jianwei Ma and GerlindPlonka, The Curvelet Transform - A review of recent applications, IEEE
SIGNAL PROCESSING MAGAZINE [118] MARCH 2010.

You might also like