The Study On Epipolar Rectification For Assisting In Taking 3D Video Programs

The 3D video program is popular in recent years as a new medium that can increase audience's viewing experience. Compared with traditional 2D video programs, it can provide a new spatial dimension making audience feel more realistic and have stronger live feeling, which greatly increases people's visual impression.Through a vast media promotion, the 3D market has grown rapidly and the concept of 3D television has been widely known. However, because of the lack of 3D film sources, the 3D television has not been popularized all over the world. When people take 3D video programs, the left and right camera can't meet the standard parallel configuration that will lead to the corresponding scan lines cross over, causing the viewer dizziness, eye swelling and visual fatigue. Therefore, making a thorough study on the epipolar rectification technology involved in 3D video programs shooting and post-process, which addresseses the misalignment of scan lines, is of great significance. Currently, the epipolar rectification technology of 3D video programs can be divided into two categories: camera calibration and no camera calibration.Under the camera calibration, the traditional scheme is to use calibration template to calibrate the left and right cameras respectively in order to get their each intrinsic and extrinsic parameters and then conduct epipolar rectification according to the perspective projection transformation. The traditional camera calibration methods focus on calibrating not only the single camera's intrinsic parameters of radial and tangential distortion but also extrinsic parameters of rotation and translation relative to the chessboard while epipolar rectification techniques emphasize the adjustment of the relative rotation and translation of the left and right camera. So the traditional camera calibration methods usually can not effectively solve the misalignment problem of scan lines. According to the requirements of epipolar rectification, this paper proposes the stereo camera calibration method which is combined with the rectification error. And the experimental results show that this method not only effectively solved the scan lines problem, but also greatly improved the viewing comfort of 3D video programs. And these camera parameters acquired by above method can provide accurate camera adjustment information to the 3D video program shooting process, making the left and right cameras to be on the standard parallel configuration. Based on this method, this paper systematically proposes a epipolar rectification calibration scheme which is feasible in actual application in the case of binocular stereo calibration and developes the corresponding FCP Plug-In .As the camera calibration process is generally complicated, and in reality the calibration parameters of many 3D videos are not readily available, therefore the epipolar rectification without camera calibration has greater flexibility.In the case of no camera calibration, the prerequisite of epipolar rectification is to know series of coordinates of corresponding points between left and right views, making the use of the restriction of the minimum distance and parallax of corresponding points to solve the fundamental matrix. Then through the projection transformation to project the left and right view onto a public space plane that is parallel to the camera baseline, it makes the corresponding points lie in the same horizontal scan line, which can eliminate the vertical parallax of 3D video program and enhance audience's viewing experience. With great flexibility, the epipolar rectification without camera calibration is an important research field in stereo vision. However, such methods rely on the exact accurate solutions of the corresponding points and the fundamental matrix. If there are mismatches in corresponding points, the estimation of the fundamental matrix will be extremely unstable, leading to a decline in the accuracy of the epipolar rectification and serious distortion of the images. Meanwhile, such epipolar rectification methods only apply to the post-processing of 3D video program and not able to provide guidance information to the camera adjustment in 3D program shooting. This paper presents a epipolar rectification method, which can be used not only in the post-processing of 3D video program but also to assist in camera adjustment. This method eliminates the mismatches in corresponding points to get the largest consistent set of points by the use of genetic consistency estimation (GCE) method, parameterizes the fundamental matrix into cameras'pose information, and establishes the camera parameters by using the least square method to minimize the error equation constructed by corresponding points. The camera parameters can not only solve the epipolar rectification matrix directly, but also can provide the information about how to adjust the cameras in 3D shooting process. Based on this method, this paper also systematically proposes a epipolar rectification calibration scheme which is feasible in actual application in the case of no stereo calibration and developes the corresponding FCP plug-in .