| Abstract: |
This paper presents a context-based predictive coding method for lossless compression of video. For this method, we propose a model to estimate level of activity in the prediction context of a pixel. This is measured in terms of slope and the same is optimally classified to results in a small number of slope bins. After finding the slope bins, we propose a LS based method to find switched predictors to be associated with the various bins. The set of the predictors are found on a frame-by-frame basis and when it is incorporated in CALIC frame work, the proposed method results in, on an average, a better compression performance than that is obtained using recently published methods - LOPT and M-CALIC. The proposed codec has higher coding complexity but much lower decoding complexity, which is necessary for real-time video decoding. The proposed method of coding, however, has much lower complexity as compared to the LOPT method, which has same order of high coding and decoding complexity. |
