Improvement of stereo matching algorithm based on sum of gradient magnitude differences and semi-global method with refinement step

A new stereo matching algorithm which uses improved matching cost computation and optimisation using the semi-global method (SGM) is proposed. The absolute difference is sensitive to low textured regions and high noise on the stereo images with radiometric distortions. To get over these problems, sum of gradient magnitude differences has been introduced at the first stage. This method is strong against the radiometric differences on the stereo images. Hence, this approach will reduce the error of preliminary data for stereo corresponding process. The SGM is used at the aggregation, and optimisation stage uses 16 different directions of 2D path. Additionally, the iterative guided filter is utilised at the refinement stage which minimises the errors and increases the accuracy. The proposed work produces accurate results and performs much better compared with some established algorithms based on the standard stereo benchmarking evaluation from the Middlebury and KITTI.