Lijun Li

Hierarchical Image Coding Matched to Unequal Error Protection Rate Compatible Punctured Convolutional Codes

Abstract:

With increasing image transmission applications used in Internet and wireless communication, the protection of encoded image data over noisy channels has become more important than before. Efficient algorithms are need to achieve a close to optimum rate-distortion bound in image channel coding performance.

In this thesis, we propose a new approach to protect encoded image data over noisy channels. We implement a two-dimensional Unequal Error Protection (UEP) image channel coding algorithm matched to the unequal error sensitivity of the output data of an Embedded Zerotree Wavelet (EZW) image encoder. Our system is implemented using one of the most competitive EZW algorithms, Set Partitioning in Hierarchical Trees (SPIHT). It can also be extended to other EZW algorithms.

In the thesis, we simulate our proposed system over a Binary Symmetric Channel (BSC), a Gaussian channel and a flat Rayleigh fading channel using Lena and Goldhill images. In all cases, the simulation results show that the two-dimensional UEP channel coding scheme can provide improvements over Equal Error Protection (EEP). It is also noted that this system improves over the EEP codes in more noisy environments. We also simulate our UEP scheme using a product code system.