Wavelet Video Compression

Fakeh, Rohmad (2006) Wavelet Video Compression. PhD thesis, Universiti Putra Malaysia.

[img] PDF
207Kb

Abstract

The thesis proposes the current approaches in wavelet technology which has provided efficient framework of multi-resolution space-frequency representation with promising applications in video processing. This discrete wavelet transform (DWT) is becoming increasingly important in visual applications because of its flexibility in representing nonstationary signals such as images and video sequences. The main objective of this thesis is to develop a wavelet video compression system. There are however many parameters within a wavelet analysis and synthesis which govern the quality of a decoded video sequences such as boundary policies, quantization threshold, decomposition strategies and the choice of wavelet filter-banks. An evaluation of the visual quality of images and video sequences at different parameter settings leads to recommendations on the wavelet filter parameters to be used in video compression. In this thesis the video compression schemes of 2D frame by frame and 3D spatio-temporal wavelet transformation are proposed. The standard spatio-temporal scheme has fixed number of sub-bands generated after the temporal decomposition and adopting adaptive quantization to the fixed number of sub-bands. The proposed spatio-temporal scheme proposed a flexible number of sub-bands generated depending on the penetration depth of the wavelet transformation. The global and level-dependent-threshold quantization methodology with the statistical adaptive estimation of wavelet shrinkage for the transformed coefficients have been adopted and provides high compression performance even without entropy coding and comparable to other coding scheme utilizing other quantization methods. The leveldependent- threshold is found to be a useful tool for providing fixed rate and used throughout the simulations for the empirical evaluation of the tested parameters as compared to global threshold. Extensive experimental investigations on a wide variety of monochrome and color images, and video sequences in QCIF, CIF and SIF resolutions are reported in this thesis. The international benchmark of visual quality evaluation of Mean Squared Error (MSE), Compression Ratio (CR), and Peak Signal to Noise Ratio (PSNR) are used as the objective measure of performance quality. Experimental results had shown that bi-orthogonal 9/7 (Bior-4.4) wavelet filters perform comparable for images and video sequences with less temporal activity however filter-banks from Symlet family (Sym-5) has shown to perform the best and out-performed others when applied to video sequences with even higher background activity such as the Car-phone and Akiyo sequences. Coding performance has been reported and performed best with dyadic DWT decomposition, periodic extensions and level-dependent threshold quantization.

Item Type:Thesis (PhD)
Subject:Video compression.
Subject:Wavelets (Mathematics).
Chairman Supervisor:Associate Professor Abdul Azim Bin Abd. Ghani, PhD
Call Number:FSKTM 2006 20
Faculty or Institute:Faculty of Computer Science and Information Technology
ID Code:438
Deposited By: Yusfauhannum Mohd Yunus
Deposited On:13 Oct 2008 13:54
Last Modified:27 May 2013 06:48

Repository Staff Only: Edit item detail

Document Download Statistics

This item has been downloaded for since 13 Oct 2008 13:54.

View statistics for "Wavelet Video Compression"


Universiti Putra Malaysia Institutional Repository

Universiti Putra Malaysia Institutional Repository is an on-line digital archive that serves as a central collection and storage of scientific information and research at the Universiti Putra Malaysia.

Currently, the collections deposited in the IR consists of Master and PhD theses, Master and PhD Project Report, Journal Articles, Journal Bulletins, Conference Papers, UPM News, Newspaper Cuttings, Patents and Inaugural Lectures.

As the policy of the university does not permit users to view thesis in full text, access is only given to the first 24 pages only.