Parallel digital image compression system for exploiting zerotree redundancies in wavelet coefficients

摘要

An image compression system for implementing a zerotree wavelet compression algorithm. The compression system uses a wavelet based coding system which takes advantage of the correlation between insignificant coefficients at different scales. The compression system uses parallel arrays of processing elements which include single instruction multiple data processors in its encoder and decoder to increase throughput and scalability. A single instruction decoder forces each processing element of the array to perform exactly the same operation on its data as the other processor elements of the array are performing on their data. If an operation must be performed only by some of the processing elements of the array, then those processing elements not performing the operation are turned off for one instruction cycle. The algorithm of the compression system is partitioned into two major components which are the wavelet transform and zerotree computations. In the encoder the wavelet transform and zerotree computations are performed on the single instruction multiple data processor array while an output processor performs lossless arithmetic coding on the symbols output from the array. Partitioning the algorithm allows the encoder to achieve an enhanced rate-distortion performance with increased throughput since computationally complex operations are performed in parallel. The decoder also achieves increased throughput since the computationally complex operations are performed in parallel.