An Efficient Global Matching Technique for Stereo Matching

Stereo matching method is a difficult issue in the computer vision. The efficiency of the stereo matching is greatly reduced by illumination difference and occlusion. A stereo matching technique using HSV color and improvised NCC measure is presented in this paper. During the cost computation step, the image is initially transformed from RGB color to HSV. Later, to tackle the drawbacks of the traditional NCC measure and enhance the matching accuracy, an improvised NCC method that is based on neighboring weighting is proposed. The dynamic programming is applied for disparity optimization. Further the disparity map generated is post processed using disparity refinement methods. The results conclude that the method proposed reduces the effect of illumination variations and also achieves a robust matching.