Structure runlength coding, briefly SRC, is an efficient coding for simple synthetic images. In the coding, only one of four symbols is substituted for each pixel. At the same time, additional 24 bits stand for the color of a pixel which has no relation to its neighbors. With the increase of colors, however, those pixels with no relation to their neighbors increase fast. It decreases dramatically the efficiency of SRC. In this paper, we suggest to separate the symbol data and color data. The symbol data can be compressed by dictionary-based coding, and color data are processed by predictive DPCM (differential pulse code modulation). The predictive coefficients are adapted by the result of SRC. The proposed method can reach the higher efficiency than the SRC, especially for images with many colors, at the cost of additional buffers needed to store all compressed symbol data and color data in a decoder.
抄録全体を表示