Abstract
In this thesis, we propose a novel algorithm to generate a virtual-view video from a video-plus-depth sequence. How to synthesize the realistic content to fill in the disocclusion regions at the synthesized view is the main challenging problem in this task. The proposed method enforces the spatial and temporal consistency in the disocclusion regions by formulating the problem as an energy minimization problem in a Markov random fields (MRFs) framework. The resulting MRF optimization problem is solved via the belief propagation (BP) algorithm. We first recover the depth images and the motion vector maps after the image warping with the depth map. Then we formulate the energy function for the MRF with additional shift variables for each node. To reduce the high computational complexity of applying BP to this problem, we present a multi-level BPs by using BP with smaller numbers of label candidates for each level. Finally, the Poisson image reconstruction is applied to improve the color consistency between the boundary of the disocclusion region in the synthesized image. Some experimental results of applying the proposed algorithm to real video-plus-depth sequences are shown to demonstrate its performance.