Year

2004

Degree Name

Doctor of Philosophy (PhD)

Department

School of Electrical, Computer and Telecommunications Engineering - Faculty of Informatics

Abstract

Rapid advances in networking technologies and the explosive growth of the Internet in recent years have increasingly raised the heterogeneity issue (in end-users processing capabilities and access bandwidth) for transmission of visual information. The traditional approach of providing multiple bitstreams, each tailored for one class of decoding requirements to address the heterogeneity issue is very inefficient and only covers very limited classes of end-users. Scalable coding that generates a bitstream which consists of a set of embedded parts that offer increasingly better signal-to-noise ratio (SNR) and/or spatial and/or temporal resolution, is widely considered as a promising coding approach for image/video transmission in heterogeneous environments. This thesis presents a family of highly scalable wavelet-based image and video coding systems based on the powerful set partitioning in hierarchical trees (SPIHT) algorithm. A framework for highly scalable SPIHT (HS-SPIHT) coding is first introduced. The main idea is to separately encode each level of the spatial resolution subbands in the wavelet decomposed image to achieve spatial scalability, while benefiting from the tree structure and the efficient set partitioning rule defined by SPIHT to preserve the compression efficiency. In this framework two novel algorithms, HS-SPIHT-I and HS-SPIHT-II, are proposed. Both algorithms fully support SNR and spatial scalability for image coding and output flexible bitstreams that are easily parsable (reorderable) for providing full embedded subbitstreams for lower resolutions. It is shown that the HS-SPIHT approach does not sacrifice compression efficiency for providing scalability features. The HS-SPIHT coding approach is extended to 3-D (3DHS-SPIHT) to provide combined SNR, spatial and temporal scalability support for video coding. An accurate motion compensation temporal filtering (MCTF) scheme is employed in the proposed highly scalable video coding system to achieve not only more efficient compression but also to produce clear, blur-free sequences for lower temporal resolutions. Simulation results prove the compression efficiency and full scalability support of the proposed video coding system. An object-based modification of HS-SPIHT (OBHS-SPIHT) for highly scalable texture coding of arbitrarily shaped image objects is also presented. It effectively encodes only the decomposed object texture, while it keeps the full scalability functionality of the HS-SPIHT. A non expansive shape adaptive discrete wavelet transform (SA-DWT) is utilized for decomposing the texture information of the object. The OBHS-SPIHT is also extended to 3-D for video object coding. The compression efficiency of the OBHS-SPIHT approach in comparison to the state-of-the-art object-based coding algorithms, as well as its full scalability feature, are shown in the simulation results. Full scalability support and object-based functionality of the proposed highly scalable image and video coding approaches with their compression efficiency and low complexity (which makes them very convenient for software implementation), make them attractive for many multimedia applications. This is especially so for multicasting visual information over heterogenous networks and object-based visual information storage, processing and retrieval systems.

Share

COinS
 

Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.