Multi-view video system has been receiving more and more attention since its applications such as three-dimensional television (3DTV) or free viewpoint television (FTV) provide customers with high interactivity and photo-realistic image quality. In this thesis, we mainly worked on three problems to be solved in realizing Multi-view video system - rectification, free view generation and color correction....[ Read more ]

Multi-view video system has been receiving more and more attention since its applications such as three-dimensional television (3DTV) or free viewpoint television (FTV) provide customers with high interactivity and photo-realistic image quality. In this thesis, we mainly worked on three problems to be solved in realizing Multi-view video system - rectification, free view generation and color correction.

Specifically, a novel ray-space based rectification algorithm is first proposed to rectify the epipolar plane images (EPIs) such that pixels corresponding to the same scene point will form a straight line in the EPI. After rectification, a novel radon based ray space interpolation algorithm is proposed to generate virtual viewpoint images. In our proposed algorithm, Radon transform is applied to detect the candidate interpolation direction set. The occlusion problem is also alleviated by our improved block matching based interpolation method in which smoothness property of the disparity field and the correlation among neighboring EPIs are explored. Experimental results show that the proposed ray-space based rectification algorithm and the radon based ray-space interpolation algorithm demonstrates quite good performance. Moreover, our proposed algorithm does not need any scene depth information.

Furthermore, since color inconsistency among different views greatly reduces the coding efficiency and rendering quality of multi-view video, a novel ray-space based color correction algorithm is proposed. Radon transform is applied in ray-space domain to detect the corresponding pixels in different viewpoint images. For each viewpoint image, the optimal color correction matrix is calculated by minimizing the error energy between the color of the current view and the average color based on the least square error criteria. Experimental results show that the proposed algorithm greatly improves the color consistency among different views. Moreover, we also demonstrate that the coding efficiency of Multi-view video is greatly improved after applying our proposed color correction algorithm.