Research on Lossless Predictive Coding Methods for Stereo Images

Abstract

There exists very high correlation between left-view images and right-view images of stereo image pairs. The correlation can help to reduce the stereoscopic image coding bit rate. Proposed methods first estimate the disparity by using causal neighborhood pixels. And then a prediction method which combines inter-view prediction with intra-view prediction is used to obtain a better predictive coding method for target-view images. The reference-view images are then encoded by using those commonly used prediction coding method. The experimental results show that compared with those methods using only intra-view correlation, the proposed method can obtain lower code rate.