Maximum-likelihood versus maximum a posteriori based local illumination and color correction algorithm for multi-view video

In multi-view video, illumination and color inconsistency among different views always exist because of imperfect camera calibration, CCD noise, camera positions and orientations, etc. Since illumination and color inconsistency greatly reduce the coding efficiency and rendering quality of multiview video, effective illumination and color correction modules are necessary for practical multi-view video processing system. In this paper, we proposed two local illumination and color correction algorithms. In these two algorithms, the correction matrix is estimated by applying maximum likelihood (ML) and maximum a posteriori (MAP) methods respectively. According to the Bayes rule, the MAP estimate is determined by two terms: error conditional density model (likelihood model)and priori conditional density model. Experimental results show that both the ML and MAP based correction matrices greatly improve the illumination and color consistency among different views. Moreover, images corrected by MAP based correction matrix look much nicer than those corrected by ML based correction matrix.