In this paper, we propose a context-based adaptive predictor for use in lossless image coding. Most often, lossless image coders utilize non-adaptive linear predictors for the sake of simplicity and to reduce the complexity of the coder. In DPCM-based lossless image coders, adaptivity can result in significant improvements in the performance. However, adaptive prediction is faced with a number of problems chiefly its extensive computational demands. The predictor proposed in this paper allows for a lower computational cost while guaranteeing the stability of the predictor. The context-based adaptive predictor (CBAP) was found to outperform or at least perform equally as well as the optimum linear predictor for a variety of test images. We should also note that designing an optimum linear predictor requires some knowledge of the image prior to coding while the CBAP requires no such knowledge and operates "on-the-fly".
展开▼
