A new arithmetic coding model for a block-based lossless image compression based on exploiting inter-block correlation

In this paper, we investigate a new approach for a block-based lossless image compression using arithmetic coding. The conventional arithmetic encoders encode and decode images pixel by pixel in raster scan order by using a statistical model which provides probabilities for the whole source symbols to be encoded. However, in the proposed scheme, the arithmetic encoders encode an image block by block from left to right, and block-row by block-row from top to bottom. The proposed model estimates the probability distribution of each block by exploiting the high correlation between neighboring image blocks. Therefore, the probability distribution of each block of pixels is estimated by minimizing the Kullback-Leibler distance between the exact probability distribution of that block and the probability distributions of its neighboring blocks in causal order. The results of comparative experiments show significant improvements over conventional arithmetic encoders in both static and adaptive order-0 models, reducing the bitrate by an average of 15.5 and 16.4 % respectively.