Stereo matching method based on image segmentation block boundary guidance and constraint

Abstract

The invention discloses a stereo matching method based on image segmentation block boundary guidance and constraint, and the method comprises the steps: segmenting an image into different regions through an image segmentation technology, and carrying out the boundary guidance and constraint of the image segmentation block. And cost calculation and constraint cost aggregation processes are guided by using boundary information of the segmentation blocks. Specifically, in a cost calculation stage, two kinds of improved Sobel operators are provided, and gradient information, color information and Census values of boundary pixels and non-boundary pixels of segmentation blocks are fused adaptively; in the cost aggregation stage, the boundary of the segmentation block is used as a constraint, the arm length of the cross arm is adaptively limited, and cost aggregation is carried out. Experimental results show that the mean absolute error of the pixels of the method is 10.5%, and the method is excellent in the aspects of keeping image edge details and improving matching precision.

Images