Fuzzy logic-based disparity selection using multiple data costs for stereo correspondence

Authors: AKHIL APPU SHETTY, V.I GEORGE, C GURUDAS NAYAK, RAVIRAJ SHETTY

Abstract: Stereo matching algorithms are capable of generating depth maps from two images of the same scene taken simultaneously from two different viewpoints. Traditionally, a single cost function is used to calculate the disparity between corresponding pixels in the left and right images. In the present research, we have considered a combination of simple data costs. A new method to combine multiple data costs is presented and a fuzzy-based disparity selection method is proposed. Experiments with different combinations of parameters are conducted and compared through the Middlebury and Kitti Stereo Vision Benchmark.

Keywords: Fuzzy logic, stereo matching, mutual information, normalized cross-correlation, Middlebury stereo dataset, Kitti stereo dataset

Full Text: PDF