Stereo Rectification Methods For Stereoscopic Video

3D movies have become very popular in recent years. But there is vertical disparities between left and right views of3D frames due to the lack of accuracy of the mechanical alignment during the shooting. Vertical disparities is one of reasons that will make the viewer feel dizzy, eye strain and visual fatigue. Therefore, the research to find the theory and method to eliminate the vertical disparities is meaningful. The process of eliminating the vertical disparities is also known as " stereo rectification ".This paper proposes a new method of stereo rectification. In order to make the epipolar lines to be parallel with the horizontal scan lines, the traditional way is to construct a mapping matrix to map the epipoles to (1,0,0)T directly, which usually causes great distortion to the image afterwards. To solve this problem, our method conduct it in two steps:1) mapping the epipoles to some points at infinity;2) aligning the epipolar lines with x-axis. In addition, by taking advantage of the fact that the stereoscopic movies are almost aligned and without great vertical disparities, our method keeps one of the stereo images unchanged, and the rectification only imposes on the other image to reduce the unnecessary distortion from transformation. The depth information is also an important attribute of the stereoscopic movie. We propose a new constraint to ensure the consistency of the disparities.The above method assumes that the camera model is the perspective camera model, namely there is no lens distortion. When there exists lens distortion in stereo video, the result will be seriously affected. The paper puts forward a solution to this problem. The distorted images have to be un-distorted to perspective images. Firstly, Fitzgibbon’s method is used to do simultaneous estimation of the fundamental matrix and distortion coefficient; Secondly, the images are un-distorted; Finally, based on the basic matrix obtained before,the transformation matrix can be estimated to do stereo rectification. The experiments of simulation and fisheye images verify the effectiveness of the proposed method.