Scalable image compression for database storage and transmission applications.

Abstract

In this thesis, we first define the concept of scalability and describe its three types, namely SNR, spatial and temporal scalability. This is followed by a review of the various techniques reported in the literature for achieving scalable image compression. We then investigate the algorithms for scalable image compression within the framework of JPEG standard. An extension to the hierarchical mode algorithm of JPEG is presented which provides an improved coding performance. This is followed by evaluation of the performance of these algorithms for image transmission over ATM networks. JPEG provides an efficient technique for achieving SNR scalability. However, JPEG algorithms do not provide a good coding performance for achieving spatial scalability at low bit rates. We propose a high performance algorithm based on wavelet transform and vector quantization for achieving spatial scalability. This algorithm ensures partial decodability of VQ labels by using multiple codebooks for image reconstruction, one for each spatial resolution. Our simulations indicate that the proposed algorithm provides spatial scalability at lower bit rates compared to the hierarchical mode of JPEG. (Abstract shortened by UMI.)