Research On Key Technologies Of Super Multi-View3D Video

Stereo vision is a hot research topic in the field of computer vision and3D video display. Getting disparity maps of high quality is one of the most crucial steps. In order to obtain correct disparity maps for3D reconstruction and computer vision, many algorithms have been developed for stereo disparity estimation in recent years. Actually, diversity of the algorithms mostly relies on the optimization approach which determines the main characteristics including complexity and precision of the methods. Balancing the accuracy and execution time is the fundamental constraints of stereo matching algorithms, especially in consumer electronics application.Here, a novel constant computational complexity algorithm based on separable successive weight summation (SWS) among horizontal and vertical direction is presented. The cost aggregation is the most important processing step to achieve more accurate disparity maps than original algorithm. The proposed algorithm eliminates iteration and its support area is independent, which saves computation and memory space compared with the SWS aggregating cost values effectively along large regions by four passes. The weight area of four directions is determined by the intensity similarity of four neighborhood pixels. The computation includes six additions and four multi-applications per pixel for each candidate disparity value. The similar measure of gradient is also applied to improve the original algorithm. The edge of image is an important characteristic, which is discontinuous area including the useful information for stereo matching. Image segmentation and edge detection is applied to the stereo matching ideas to accelerate the speed and accuracy of matching algorithm. The image edge is extracted to reduce search scope for the stereo matching algorithm. In the discontinuity region of the image edge, search scope is increased to acquire better matching effect. Dense disparity map is obtained through local optimization. To test the performance of the stereo matching algorithm, it is realized by C++. Experimental results show that the algorithm is fast and efficient, and the matching noise is well reduced and the matching precision is improved in depth discontinuities and low-texture regions.This paper proposes a solution to synthesize and display multi-view stereo video with32viewpoints. While inputting a2D video and its depth maps, this solution allow generating a32viewpoint video and outputting to the naked eye stereo display devices though some image processing procedures such as depth maps extraction, DIBR algorithm, cavity filling and so on. Looking in front of the display screen with appropriate location, we can find that the stereoscopic effect of displaying32viewpoint videos is obvious with no obvious operation and display error. From the experimental results we can see that this solution can achieve the expected design goal.