Improved Census Transform Method for Semi-Global Matching Algorithm
Xie Pan, Guoben Jun, Yuanping Xu, Zhijie Xu, Tukun Li, Jian Huang, Wenbo Qiao · 2021
In order to solve the problem that the original Census transform algorithm is easily affected by the gray value of the central pixel, and further improvement of the matching accuracy of SGM algorithm, this study proposes an improved Census transform algorithm based on the sequence template. It divides the pixel values in the Census window into sequential sequences from the outside to inside, and then constructs Census encoding templates within and between sequences according to predefined rules in this study, such that the Census value of the center pixel can contain more information within the window, and can further improve the matching accuracy of SGM. Experimental results show that the improved sequence template-based Census transform method can enhance average 16.26% matching accuracy of SGM algorithm, while preserving its efficiency.