Stereo Image Correction And Video Encoding

Stereoscopic or in general multi-view video can provide more vivid an accurate information about the scene structure than from mono-view video. However one major obstacle for using multi-view video is the extremely large amount of data associated with it. This dissertation considers the problem in multi-view video system, including video preprocessing and video coding.First, we describe an image rectify approach, which can convert images captured using non-parallel cameras to coplanar like images. This approach greatly eases the computational burden incurred by the non-parallel camera geometry, where one must consider both horizontal and vertical disparities. Also, we introduce an approach for intermediate view generation from stereo pair acquired by two parallel cameras.Second, a luminance and chrominance correction algorithm is proposed to deal with the significant luminance and chrominance discrepancies between the different camera views. According to the theory of image formation, Luminance and Chrominance variation between the images of different cameras were categorized into multiplicative and additive errors. Histogram matching is used to compute multiplicative and additive variation factors for its low computational complexity.Finally we present a coding scheme based on H.264 for stereo video. The proposed coding are based on variable size blocks using both motion and disparity estimation. Furthermore RD-based mode selection is adopted to manage the increasing mode introduced by the variable size blocks based disparity estimation.