Stereo matching algorithm based on improved Census transform and path cost

Abstract

In order to reduce the mismatching rate and improve the accuracy of disparity calculation, a stereo matching algorithm base on improved Census transform and path cost was proposed. Firstly, a reference pixel is applied to replace the center pixel by setting a threshold to improve the anti-noise ability. The improved Census transform algorithm is normalized weighted fusion with absolute differences (AD) algorithm and gradient information to improve the accuracy of parallax calculation and to enhance the stability of the algorithm. Secondly, a penalty factors determination method based on the pixel color difference on the path is designed to improve the matching effect of different regions. Finally, weighted median filtering and discontinuous disparity adjustment steps are added to further enhance the performance of the algorithm. The effectiveness of the proposed algorithm is verified by using 14 images which are selected from the Middlebury 2.0 test platform. The experimental results show that the average mismatch rate is 4.89%, and the proposed algorithm can effectively reduce the mismatching rate and improve the disparity effect.

Keywords

• stereo matching
• census transform
• normalized weighted fusion
• path cost aggregation
• color difference