Technology Research Of Image Correction And Stereo Matching In Stereo Video Collection

With the development of information technology,stereo video has gradually entered people's life,such as 3D movie and free-view TV.Stereo video system includes video capture,video codec,video transmission and display,and many other key links.Stereo video capture is at the forefront of the system,including color video sequence acquisition and deep sequence acquisition.The quality of the collected information seriously affects the subsequent processing.In order to further enhance the image quality of stereo video,improve the efficiency of stereo video acquisition,the following work is done in this paper,for image correction and stereo matching in stereo video capture link:First: The basic theory of stereo video is studied,and the existing image correction and depth estimation methods are classified and collated to analyzing their advantages and disadvantages.Second: Image correction is the key to eliminating parallax deviation and gaining high quality stereo video sequences.In this paper,an improved image correction method based on camera calibration is proposed for the stereo video sequence captured by parallel array of cameras.At the same time,the depth information of the scene is estimated by corner detection and matching to further reducing the error of correction.Through the experimental analysis,the improved method not only improves the accuracy of the correction,but also reduces the image loss rate caused by the correction.Third: Global stereo matching can obtain an accurate depth sequence,but the computational complexity is too high,and the time consistency of the the resulting depth sequence is poor.Therefore,a fast stereo matching method based on graph cutting is proposed in this paper.In this method,video sequence will be divided into groups of images with the same length.Stereo matching is used for the first frame of each group,and three-frame subtraction motion detection will be used for the other frames.Those frames will be divided into static area and sports area according to the motion detection results.Static area will directly refer to previous frame for depth information,and sports area will use stereo matching.At the same time,the traditional expansion move graph cut method is improved in this paper.The parallax label is initialized using the depth information of previous frame,and the label order will be dynamically modified,and a time smoothing item is added to the global energy function.Experimental results show that the proposed method can improve the speed of stereo video depth estimation,and enhance the time consistency of the depth sequence.