Research On Key Technologies Of Auto-stereoscopic 3D Augmented Reality

Augmented reality is a view of real-world environment whose elements are augmented by computer.3D augmented reality technology focuses on an auto-stereoscopic 3D representation of augmented reality. By applying augmented reality to Natural 3DTV, observers are able to experience an enhanced perception of 3D world and a live interaction.Natural 3DTV applies stereo shooting to get scene information and encode, transfer, ecode this information. Multiview construction module generates virtual view images and delivers stereo image pairs into left and right eyes individually to get observers natual 3D feeling. Based on exisiting Nature 3DTV platform, this paper focuses on 3D augmented reality, especially the feature detection and matching technology and the depth map refinement based on matting algorithm. Achievements are as follows:(1) An augmented reality platform based on monocular SLAM technology is built. The platform builds and updates a map with the environment and at the same time keeps track of current location of camera.(2) The paper researches on feature detection and matching technology and proposes a fast SIFT algorithm based on Sobel algorithm. As SIFT detector has a high degree of redundancy and a low speed, we apply Sobel edge detector to generate an edge-gathering spatial scale and detect the extreme point under the constraint of it. The experimental results show that the proposed algorithm decreases the redundancy of key points and speeds up the implementation while the matching rate maintaines at a high level. As the threshold of Sobel detector increases, number of keypoints decreases and matching rate gets higher.(3) A depth refinement technique based on matting algorithm is presented. Limitation of stereo correspondence causes boundary artifacts in virtual view images. With the alpha value alculated by matting algorithm, this tichinique is dedicated to recalculate the depth value of (?) pixels and refine the virtual view images. The distinguishable improvement in this paper is the localization of matting algorithm which is used to solve the multiple-layer definition in complex depth map. Experimental result suggests that this technique reduce the boundary artifacts and improve the quality of virtual view images effectively. Localization of matting algorithm shows better perfomance than general algorithm does.