Lossless Image Compression via Classified Rearrangement and Group Encoding

Abstract

Listen

Translate article

This paper proposes a lossless image compression algorithm based on classified rearrangement and group encoding.The algorithm exploits the effects of grayscale distribution of the original image on the compression ratio, and first classifies all the pixels into several classes in terms of the probability statistics of their grayscales, and then rearranges all classes of pixels along inner-block and inter-block according to Hilbert curves, increasing the correlation among pixels by reordering their spatial distribution, and last respectively does group encoding to every class of pixels on the basis of their grayscale probabilities, which makes the most of the correlation among pixels to achieve lossless compression. Group encoding assigns one-to-one codeword of variable length to each grayscale in accordance with its probability value, and the codeword is comprised of group number and inner-group representation two parts. The simulation results of multiple test images demonstrate the algorithm is of low complexity and high performance.