Color image compression based on DWT

Other Title(s)

ضغط الصورة الملونة بالاعتماد على التحويل المويجي الخطي

Dissertant

Muhammad, Faysal Ghazi

Thesis advisor

Jurj, Luayy Adwar
al-Abbudi, Bushra Qasim

University

University of Baghdad

Faculty

College of Science

Department

Astronomy and Space Department

University Country

Iraq

Degree

Ph.D.

Degree Date

2006

English Abstract

In this research work, some low complexity and efficient coding approaches are proposed to compress color images.

The suggested compression schemes consist of some coding modules.

The first coding modules is indicated to perform color transform (i.e., YUV, YIQ, or YCbCr), their suitability was investigated experimentally.

The second coding step is the application of the wavelet transform (tree decomposition); three known wavelet filters (i.e., Harr or biorthogonal wavelet filters (Tab3/5 and Tab7/9)) have been tested as the transform basis.

The wavelet transform is exploited to resample the two chrominance subbands, and used to transform the luminance subband into its approximation coefficients and wavelet (detail) coefficients.

The approximation coefficients of the two chrominance subbands and luminance subbands have been coded by using proposed lossless scheme.

The third coding step is concerned with encoding the wavelet coefficients of the luminance subband.

This is done by using a heirarical scalar quantization followed by modified bit-slicing method to select the significant wavelet coefficients to be finally encoded using a suitable lossless coder.

Huffman and LZW are used as lossless encoders.

Three methods were proposed to solve the wavelet coefficients selection problem.

The first is based on the idea of eliminating one (or more) of the least significant bit slice.

While in the second method, the vector quantization mapping techniques is used, where the code-book is build for a given block size, then the bit-slices are partitioned into nonoverlapped blocks.

Then, for each block the code-book is searched to find the best matched template, and finally the block index of the best matched template is sent instead of sending all the block elements.

This third coding method (which is the best among previous mentioned methods) is based on the fact that the most repetitive blocks of the bitplanes (especially the most significant bit-planes) are blocks whose contents are (0’s), “cluster of zeros bits”.

So, such blocks should be coded as one zero bit, while all other blocks shown should be coded by using variable codewords (whose size will be more than one bit).

The obtained results showed that the performance of lossy schemes is image dependent; wavelet filters varies in their performance, since no specific biorthogonal wavelet filter performs better than the other on all images.

Standard real images were used as test materials to investigate the performance of the suggested compression scheme; the results indicate that the efficiency of proposed scheme is encouraging when it is compared with state of the art JPEG and JPEG2000

Main Subjects

Astronomy

Topics

• Data processing

No. of Pages

107

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-604446