BTC image coding with visual patterns

This paper presents a block truncation coding (BTC) scheme for image compression that employs visual patterns designed independently of the image to be coded. In this method, image blocks are first classified into three classes: uniform, gentle edge, and sharp edge, according to their variations. Design focus is on edge blocks, where two sets of patterns, namely, visual patterns and visual row patterns are used. Different quantization is performed on the mean values of each block according to its block type. Experimental results show that the proposed BTC scheme can achieve PSNR about 30 dB at bit rate 0.46/spl sim/0.53 bpp (for the image Lenna of size 512/spl times/512).