원문정보
초록
영어
Due to a high percentage of the bad pixels in small baseline stereo matching method, a small baseline stereo matching method based on adaptive weight is proposed. Firstly, the size of reference window is adaptively calculated for each reference point and then matching costs is evaluated according to adaptive weights and reference window size. Secondly, winner take all is used to evaluate initial disparities and unreliable matches are refused by using reliability constraints. Finally, disparity post processing method based on iterative diffuse using a new cost function is used to obtain a dense disparity map. A small baseline stereo image pair of Toulouse and stereo image pairs provided by the benchmark Middlebury database are used to test the proposed method. The experimental results show that this method can effectively reduce a percentage of the bad pixels of depth discontinuity areas, improve the overall disparity matching accuracy and meet the requirements of small baseline three-dimensional reconstruction.
목차
1. Introduction
2. Fundamental Principles
3. Key Steps of the Proposed Algorithm
3.1. Adaptive Support-Weight
3.2. Adaptive Window
3.3. Reliable Constraints
3.4. Disparity Post Processing
4. Experimental Results
4.1. Experimental Environments
4.2. Accuracy Analysis
5. Conclusion and Future Work
Acknowledgements
References