The Application And Research Of Error-Resilient In Image Coding

As we know, in wireless communication, wireless channels always produce outburst errors , so it will cause high BER in wireless communication system, and in wire communication, network always produce congestion or errors, so it will cause packet loss, time-out, bit transmission errors, bit block errors etc, when image code stream are being transmitted. After the coded image having been transmitted through the channel, based on the factors talked above, the quality of reconstructed image will be bound to drop dramatically. This demands us to study image error-resilient coding algorithm to ensure the quality of reconstructed image to be good. Once a error appears in coding systems of error-resilient characteristic, if only not a fatal error, it needn't to be corrected. Especially in image coding, a few coefficient errors won't influence vision effect much. Error-resilient's main purpose is to build up robust of the system so that the image coding/decoding system can be all right under a few errors.In this paper, we introduce some classical coding methods and modern coding algorithm firstly, Secondly, we discuss and analyze some excellent error-resilient algorithm today, such as: multiple description image coding based on wavelet transforms, error-resilient pyramid vector quantization for image compression, robust wavelet zerotree image compression with fixed length packetization, multistage vector quantization codebook design for wireless image communication using genetic algorithms and JPEG2000 coding standard.Then, we present the virtual block error-resilient algorithm based on SBC. The algorithm will be implemented in two steps, the first step is to set up the frame of SBC algorithm: Firstly, we divides the coefficients of integer-to-integer wavelet transform into bit planes within each subband which are named as subband bit-planes. For these subband bit-planes, the algorithm uses simple but effective rate-distortion optimization method to determine the coding order. According to the R-D optimized coding order, the coefficient bits are coded by adaptive MQ arithmetic coder. After we have adopted SBC algorithm, those significant bits which play a key part in reconstructing image are only a very little portion compared to all the bits, so the probability of significant bits being "polluted" is reduced. This means that a majority of all the bits are non-significant bits, but their influence on reconstructing image is very little, even though there are errors in them, the quality of reconstructed image will reduce little; the second step is to add virtual block error-resilient algorithm: firstly, we divide all the image into several equalblocks, secondly, we apply wavelet transforms on the whole image (before the image is divided into several equal blocks), after wavelet transforms has been taken, we divide wavelet coefficient into correspond blocks according to image's being divided into several equal blocks, then we adopt SBC algorithm to code all the block's coefficient; At the same rime, we add a synchronization flag so that each block which has been compressed can be decoded independently. Even though errors occur in blocks, the errors will be restricted in blocks where errors occur, the next block will be decoded correctly in the affect of synchronization flag, so, the algorithm's ability of error-resilient will be enhanced further. Based on the research we have done before, we design and implement virtual block error-resilient system based on SBC algorithm.In the end of the paper, we have done a great deal of experiments using many standard test images. Compare our algorism to those fine error-resilient algorithm such as PZW, ET-SPIHT\ JPEG2000 etc, we come to the conclusion that at different compression ratio, our error-resilient algorithm are stronger in the whole(the PSNR of reconstructed image increase ldB-2dB)and more steady; In different BER, our algorithm is also more robust than the others (the PSNR of reconstructed image increase 2dB-4dB), and higher the BER is, more excellent our algorithm w...