A new, error-tolerant, still image compression algorithm is described. An error-resilient adaptation of an existing zerotree wavelet coder is used to compress the image. Then, a variable-rate, concatenated error correction code is applied to protect the data further. Variable-rate protection exploits the progressive coding ability of the compressor, encouraging uncorrected errors to appear in the least significant portions of the bit-stream during decoding, which lessens their effect on the final image. The error-resilient organisation of the information inhibits the propagation of uncorrected errors, ensuring graceful degradation of the image over a wide range of bit error rates. Results for the popular 512 x 512 'Lena' image are presented, showing notable improvements over previously published algorithms.
|Number of pages||5|
|Journal||IEE Proceedings - Vision, Image & Signal Processing|
|Publication status||Published - 1 Jan 2001|