Stereo image coding with disparity estimation using maximum a posteriori (MAP) algorithm

摘要

This paper derived the generalized formula of disparity estimation based on Bayesian Maximum A Posteriori (MAP) algorithm, and proposed the probabilistic plane configuration model which consisted of independence and similarity among the neighboring disparities in the configuration. The paper utilized the unequal distribution of frequency components, or one directional errors in the residuals of stereogram, which resulted from one dimensional epipolar geometry. The adaptive quantization and scanning scheme were compared with the conventional coders. In addition, the paper proposed a hierarchical scheme of disparity compensation in the application of multiple-view stereo images. According to the experiments, the derived formula and the proposed probabilistic models had good performances, and the hierarchical scheme of disparity compensation with multiple-view stereos improved the performance without any additional overhead to the decoder. In comparing with the coding results, the adaptive quantization and reordered scanning to utilize the directional errors showed better compression performances than the usual quantization and zigzag scanning did.