Accurate Disparity Map Estimation Based on Edge-preserving Filter

This paper proposes a new algorithm to estimate a disparity map. This map contains depth information from stereo matching process. Generally, this process is sensitive to low texture areas and high noise on images with high different brightness and contrast. To get over these problems, the proposed algorithm utilizes the RGB channels at the matching stage and edge-preserving filter at the second and final stage. The filter is known as Guided Filter (GF). The GF kernel well-recovered low texture areas which is able to reduce noise and sharpen the images. Additionally, GF is strong against the distortions due to high brightness and contrast. The propose algorithm produces accurate results on the disparity map for the low textured regions. The proposed work in this paper produces accurate results and performs much better compared to some established algorithms based on the quantitative and qualitative measurements using standard stereo benchmarking evaluation from the Middlebury.