Depth information is one of the most important elements in generating three-dimensional (3D) content. Stereo matching methods estimate depth information using the binocular characteristic. The estimated depth information is typically represented by a disparity value. Therefore, two slightly different viewpoints are used to find the disparity value. However, in the homogeneous region, corresponding point finding is problematic since the area is textureless. In order to solve this problem, we propose a pixel based cost computation method using weighted distance information for cross-scale stereo matching. The proposed method uses a hierarchical structure to accurately estimate disparity values in the homogeneous region. We also employ the distance information to complement the pixel based cost function. The experiment results show that the proposed method exceeds the conventional cross-scale stereo matching in terms of produces accurate disparity values.
Yong-Jun Chang, Yo-Sung Ho, "Pixel Based Cost Computation Using Weighted Distance Information for Cross-Scale Stereo Matching" in Proc. IS&T Int’l. Symp. on Electronic Imaging: Intelligent Robots and Computer Vision XXXIII: Algorithms and Techniques, 2016, https://doi.org/10.2352/ISSN.2470-1173.2016.10.ROBVIS-393