Efficient Binocular Stereo Matching Based on Sad and Improved Census Transformation

Abstract

Binocular stereo matching aims to obtain disparities from two very close views. Existing stereo matching methods may cause false matching when there are much image noise and disparity discontinuities. This paper proposes a novel binocular stereo matching algorithm based on SAD and improved Census transformation. We first perform improved Census transformation, and then we get the matching costs by combining SAD and improved Census transformation. Finally we cluster the matching costs and calculate the disparities. To generate better disparities, we further propose the improved bilateral and selective filters to enhance the accuracy of disparities. Experimental results show that our binocular stereo matching can produce more accurate and complete disparities, and it works well in complex scenes with irregular shapes and more objects, thus it has wide applications in stereoscopic image processing.