Calibration Technology In The System Of Auto-stereoscopic Display

With the continuous development and maturation of3D technology, theapplications of stereoscopic video are no longer confined to3D movies. Stereoscopictelevision is gradually coming into the horizon and receives extensive attention. Butthere are still a series technical problems need to be resolved in the acquisition anddisplay of stereoscopic video. Among them, the existence of vertical parallax makesstereoscopic synthetic image distortion, affecting view enjoyment stereoscopic video,seriously, and even causes severe visual fatigue. So it must to be eliminated in someway. Video Calibration technology can solve this problem very well.The paper first made a detailed study on two kinds of correction methods ofmulti-view image and experimentally verified. In the correction method based oncalibration, we use Z. Zhang plane calibration algorithm to estimate calibrationparameters which is used to design rotation matrix and to calibrate multi-cameraunder parallel configuration. This method reduces vertical parallax between theviewpoints effectively and improves horizontal parallax between adjacent cameras.But the problem of optical center collinear between multiple cameras can’t beresolved through multi-view transform matrix. In the correction method based onnon-calibrated, we extract SIFT feature points from multi-view image, using intoimage correction of four viewpoints. This method can effectively eliminate thevertical parallax, but can’t select the appropriate optical axis direction, and there is noa valid optimization way to control the distortion caused by correction and to reducethe correction errors.This paper proposes a correction method based on the parallax analysis ofmulti-view in the background of multi-camera acquisition system under parallelconfiguration. Firstly, the method selects a suitable optical axis through camerascalibration parameters, doing rotation transformation on the image plane and makingthem parallel to each other. Secondly, Secondly, we decompose the vertical disparitybetween viewpoints into two parts, which are caused by the depth inconsistent ofimages plane along the Z axis and the height difference of optical center of the cameraon the Y-axis direction, respectively. We can obtain from the experimental data thatthe vertical disparity tends to zero substantially in the system of multi-camera acquisition using the same type of cameras under parallel configuration, which can beignored. Then, we eliminate the vertical parallax with feature point matching methodand balance the horizontal parallax between, finally. The experimental results showthat this method of eliminating vertical parallax is simple and effective and the effectof correction for multi-view image is obvious, which can be applied toauto-stereoscopic display system.