Minimum spanning tree based stereo matching using image edge and brightness information
Chao He, Congxuan Zhang, Zhen Chen, Shaofeng Jiang · 2017
In this paper, we propose a MST-based stereo matching method using image edge and brightness information due to the classical MST based methods were used to produce the inaccurate matching weight in the areas of image boundaries and similar color background. Firstly, we applied the bilateral filtering to the original images to restrain the image noise and preserve the image edges. Secondly, we planned the self-adaptive weight function by combining the image edges and brightness information to improve the accuracy of the matching cost computation. Finally, we compute the disparity and refine the result according to the cost aggregation. The standard image sets of Middlebury database are employed to indicate the performances of the proposed and the NLCA method. The experimental results show the percentage of bad matching pixels of the proposed method were reduced 5.93% & 5.32% compared to those of the NLCA method with the Middlebury data. The compared results demonstrate that the proposed method has the higher matching accuracy and better robustness.