Development of Stereo Matching Algorithm Based on Guided Filter

摘要

This article presents a new stereo matching algorithm to estimate a disparity or depth map. This map contains depth information from stereo matching process. Commonly, the region of low texture area is the most difficult and challenging area for matching process to get accurate result. Thus, the proposed work in this article uses an edge preserving filter to decrease the noise and surge the accuracy on this region. The filter is known as Guided Filter (GF). The GF is robust on the low texture area with high contrast and brightness. It is also capable to sharpen and decrease the noise on the filtered images. From the standard dataset for stereo evaluation of the Middlebury, the propose framework gains precise results and perform better than some of established algorithms in the evaluated database.