Data-Efficient Hierarchical Stereo Matching Based on Variate Intensity Quantization

• Main Authors: Pin-zhih Lin, 林品芝
• Other Authors: none
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/89253466689885411020

碩士 === 國立中正大學 === 電機工程所 === 98 === Stereo matching is one of the most important topic in computer vision. The purpose is to extract the disparity or depth of the real scenes. It can then be used for 3D model reconstructon for the real world. The most popular technique for deriving the depth information of the scene is to perform the stereo matching algorithms on two or multiple images. Although the recent progress on stereo matching has improved the correctness of disparity maps, the computation is increasely dramatically, in terms of computation time and the amount of data for processing. Therefore, this thesis proposes a variate bit rate image processing which is suitable for the existing stereo matching algorithm.By combining the variate bit rate and stereo matching operation, the image information and the range of operation can be controlled during the stereo matching process amount of information for transmission and computation can be minimized. computation and increasing computing speed. The experimental results have demonstrated that the proposed framework is able to generate equivalent or better disparity maps based on the existing stereo matching techniques.