A codeword-rotation algorithm is proposed for vector quantization (VQ) of images. A novel binary classifier is presented to preclassify the training vectors into six classes including edge blocks and nonedge blocks. T...
详细信息
A codeword-rotation algorithm is proposed for vector quantization (VQ) of images. A novel binary classifier is presented to preclassify the training vectors into six classes including edge blocks and nonedge blocks. The VQ codebook is generated by applying the modified Fuzzy C-Means (MFCM) algorithm to the training vectors of each class. Similar edge blocks are rotated and coalesced during the edge subcodebook generation process. Furthermore, two schemes for designing the encoder and decoder are also presented. Compared with the basic VQ system constructed by the leg algorithm, the new method results in a considerable reduction in codebook size and computation time of codebook generation. More importantly, the visual quality achieved is better than the basic VQ system.
暂无评论