The Study And Implementation Of Multi-view Stereo Image Rectification

With the rapid development of three-dimensional(3D) technology at the 21 st century, the research on multi-view video technology is on the rise around the world. Multi-view video provide audience better feeling of three-dimension and nature, since multi-view video has multi-view information and depth message. Other more, by providing the viewer with a plurality of viewpoints, multi-view video enables us to view 3D space by freely changing our viewpoints, and to experience three-dimensional and immersive feelings. Recently, with the rapid development of free video TV(FTV) and 3DTV, multi-view video technology has attracted a wide attention. Due to the stent deformation, different intrinsic characteristics of cameras and manual adjustment errors in the producing process of multi-view video, the actually shot multi-view video cannot meet the ideal shooting quality. And these deficiencies seriously influenced the quality of stereo images produced by the back-end system. Therefore, the multi-view image rectification is of great significance in the production of multi-view video.At present, multi-view image rectification technology stay exploratory stage and there has been no one effective method. We carried out in-depth investigations in the aspects of multi-view image rectification for matrix camera arrangement in this paper. We use an ergodic method to find out the ideal rotation matrix and the ideal location of every optical center, and get the ideal projective matrix that makes the matrix camera arrangement satisfy the ideal condition. Then we calculate the image rectifying transform matrix by the homography between the original and ideal matrix camera arrangement. Experiments illustrate that the multi-view image rectification has efficient and accurate rectifying quality, and the proposed algorithm has strong robustness. Moreover the algorithm proposed in this paper is easy patulous, not only it can be directly applied on multi-view image rectification for parallel camera array but also it lays the foundation for the multi-view image rectification of other camera arrangement.