Adaptive Weight Based Sparse Block Aggregation Algorithm for Stereo Matching
Cheng Han, Wen Liu, Longbin Jin, Shan Jiang, Hua Li · 2019
To overcome the low performance in efficiency and accuracy in stereo matching, a stereo matching method that combines adaptive weight with sparse block aggregation is presented to handle the region where the depth is not continuous. Specifically, an improved matching cost function is derived from incorporating intra-color difference of pixel points into the adaptive weight to improve the success matching rate. Furthermore, a sparse mode of the interlaced columns and the center point cross direction is adopted at the cost aggregation stage, which can boost the calculation procedure. The left-right consistency detection and median filtering are utilized to rectify the mismatch points. Experimental results show that the average mismatch rate of the optimized method reduces by 5.66% in the region where the depth is not continuous and the running time of the optimized method is 15 times shorter than the ASW algorithm. Moreover, it is ultra-robust to illumination interference.