Multi-view Video Coding Based On Vision Geometry

H.264/AVC is the latest and most promising video coding standard. It is a very efficient video compression coding method which pursuits of coding efficiency and offers a very good video quality at the same time. Multi-view video captured by synchronized cameras from different viewpoints. It comprises rich 3-D information and meets people's needs of added realism and naturally reproduction of video scene. So it attracts more and more attentions in academic and industrial circle, and became the hotspot of the video area. The Major addition to H.264/AVC containing the Amendment for Multi-view Video Coding (MVC) extension, including the Multi-view High profile has been proposed by JVT in March 2009. However, Multi-view video results in a tremendous amount of data depending on the number of cameras and there are a large number of inter-viewpoint redundancy between different viewpoints. Multi-view video coding focused on compression for efficient storage and transmission of multi-view video data and how to eliminate the inter-viewpoint redundancy became the key to improve the efficiency of multi-view video coding.Epipolar geometry refers to the geometry of Computer Vision. When two cameras view a 3D scene from two distinct positions, it describes a number of geometric relations between the 3D points and their projections onto the 2D images that lead to constraints between the image points. Any two images of viewpoints of multi-view video are subject to the epipolar geometry relationship.To efficiently encode data-intensive multi-view imaging content, a fast disparity estimation search method is proposed in this paper based on the rectified epipolar lines. Firstly, a set of matching points is obtained by a DoG detection algorithm and matching of DAISY descriptor. Then a robust algorithm is used to estimate the fundamental matrix. And a specific algorithm for image rectification is proposed in this paper based on the relationship of the images of different viewpoints of multi-view video. A rectify-table which contains information of relationship between the original image and the rectified image is computed and used in the disparity estimation searching to confine the search range from 2-D to 1-D. This algorithm can reconstruct images with similar quality compared with the full search and the TZ fast search methods in JMVC (Joint Multi-view Video Coding), but the total encoding time is reduced by 30% averagely.