A new error resilience technique for image compression using arithmetic coding

Arithmetic coding is commonly used for entropy coding in image compression. It has been shown that arithmetic coding has a higher coding efficiency than other entropy coding schemes. However, it has a severe drawback when used in an error prone environment. Since arithmetic coding makes it impossible to precisely locate an error, a large portion of a bitstream must be discarded when an error occurs. In traditional packetization techniques, the bitstream is segmented to packets and the arithmetic encoder is reset at the beginning of each packet. However, the re-initialization will significantly reduce the coding efficiency of the arithmetic coder. We propose a novel technique to improve the error resilience of arithmetic coding. This new technique adds segment markers in the coded bitstream without the re-initialization of the arithmetic coder, which helps retain the correctly decoded segments within the bitstream improving the quality of the reconstructed image. Segment markers have been adopted in both MPEG-4 and JPEG 2000. Experimental results show that THE segment marker significantly improves the quality of reconstructed image with minor increase in bitstream size