Research On Local Stereo Matching Algorithm Based On Improved Census Transform

Stereo matching plays an important role in the binocular vision system.In the research of stereo matching algorithm,it is found that the Census transform stereo matching algorithm has certain robustness and fast running speed under uneven illumination and noisy environment.At the same time,the setting of the parallax search range also matches the matching algorithm.The accuracy and speed of operation have a certain influence.Therefore,this paper focuses on the Census transform stereo matching algorithm and the method of determining the range of disparity search.Aiming at the problem that the traditional Census transform stereo matching algorithm excessively relies on pixel points in generating the Census transform code and the discontinuity in the disparity range,a variable window stereo matching algorithm based on improved Census transform is proposed.In the process of Census transformation,the algorithm uses the average of the grayscale values of the pixels in the eight neighbourhood window where the central pixel is located to replace the Census transformation based on the central pixel in the traditional Census transformation,which effectively solves the traditional Census.The problem of over-reliance on the center point in the transformation improves the robustness of the algorithm in a noisy environment.At the same time,in the process of matching the pixels in the image,according to the gradient values of the horizontal and vertical directions of the pixels,a suitable matching window is selected for the points,which improves the accuracy of matching the Census transformation algorithm in the depth discontinuity region.Compared with other methods for determining the disparity search range,the image iterative subdivision method has the characteristics of high precision and high speed,but there is still the problem that the disparity search range estimation error occurs when the foreground area and the background area of the image are unevenly distributed.Therefore,this paper proposed a method for estimating the range of disparity based on forward search.In the process of image block matching,the current matching block is continued to be divided into several sub-blocks by the forward search strategy,and the parallax value of the current matching block is expressed with the matching result of its sub-block.It solves the estimation error problem caused by uneven distribution of the image block due to the background and the foreground,and further improves the accuracy of the estimation range of the disparity search range.Applying the parallax search range determination method to the improved Census transform stereo matching algorithm can improve the performance of the Census transform stereo matching algorithm.The experimental data of Tsukuba,Teddy,Venus and Cones were performed respectively.The results showed that the improved algorithm can reduce the average false-match rate of the four images by 33%and improve it further without significantly increasing the running time of the algorithm.