Translator Disclaimer
25 October 2004 Multiple wavelet-tree-based image coding and robust transmission
Author Affiliations +
Proceedings Volume 5600, Multimedia Systems and Applications VII; (2004)
Event: Optics East, 2004, Philadelphia, Pennsylvania, United States
In this paper, we present techniques based on multiple wavelet-tree coding for robust image transmission. The algorithm of set partitioning in hierarchical trees (SPIHT) is a state-of-the-art technique for image compression. This variable length coding (VLC) technique, however, is extremely sensitive to channel errors. To improve the error resilience capability and in the meantime to keep the high source coding efficiency through VLC, we propose to encode each wavelet tree or a group of wavelet trees using SPIHT algorithm independently. Instead of encoding the entire image as one bitstream, multiple bitstreams are generated. Therefore, error propagation is limited within individual bitstream. Two methods based on subsampling and human visual sensitivity are proposed to group the wavelet trees. The multiple bitstreams are further protected by the rate compatible puncture convolutional (RCPC) codes. Unequal error protection are provided for both different bitstreams and different bit segments inside each bitstream. We also investigate the improvement of error resilience through error resilient entropy coding (EREC) and wavelet tree coding when channels are slightly corruptive. A simple post-processing technique is also proposed to alleviate the effect of residual errors. We demonstrate through simulations that systems with these techniques can achieve much better performance than systems transmitting a single bitstream in noisy environments.
© (2004) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Lei Cao and Chang Wen Chen "Multiple wavelet-tree-based image coding and robust transmission", Proc. SPIE 5600, Multimedia Systems and Applications VII, (25 October 2004);

Back to Top