Abstract:
Stereo correspondence finds corresponding matching pixels in the stereo image pairs. The difference between the coordinates of these matching pixels gives the disparity value, which in turn can be used for finding the depth information of a scene.In last few decades, a number of stereo matching methods have been proposed for different Computer Vision applications, such as robot navigation, 3D modelling, object detection, tracking etc. In view of this, a new Gabor feature-based stereo matching method is proposed. Our method mainly has four steps, namely matching cost computation, cost aggregation, disparity map computation using Winner-Take-All (WTA) selection, and finally disparity map refinement. In our proposed method, local features extracted by using Gabor wavelet in spatial domain are used for matching cost computation. Subsequently, Kuwahara filter is employed for cost aggregation to smooth the estimated disparity map by preserving the disparity discontinuities.
