融合明度特征的V-SAD立体匹配算法

针对SAD算法难以对图像中纯色与弱纹理部分进行准确匹配的问题,提出了将HSV空间明度特征与SAD算法相融合的立体匹配算法,称为V-SAD算法.首先将图像从RGB空间转化至HSV空间,并根据H、S、V值将像素点按照颜色分为10类,同时得到明度特征图.然后结合SAD算法需要的灰度特征图计算匹配代价,得到初步的视差图.接着,根据HSV空间的颜色信息对图像进行分割,结合数学形态法求解分割区块中的独立连通域.再利用边缘生长法对每一个连通域的视差进行恢复.最后,使用左右一致性检测方法对视差图进行优化.实验结果表明,利用图像的明度信息衡量纯色与弱纹理区域的匹配点的相似度是有效的,该V-SAD算法有效改善了SAD算法在弱纹理和纯色区域的匹配效果,平均误匹配率下降了13.02％.

V-SAD Stereo Matching Algorithm Combining Value Features

Aiming at the problem that the SAD algorithm is difficult to match the parts with solid color and weak texture in the image effectively, this paper proposes a stereo matching algorithm called V-SAD algorithm. This algorithm fuses SAD algorithm with the value features in HSV space. Firstly, the image is converted from RGB space to HSV space, and all pixels are divided into 10 color categories according to values of H, S, and V. The value feature map can be obtained at the same time. Secondly, the matching cost is calculated combining the gray feature map required by the SAD algorithm with the value feature map to obtain a preliminary disparity map. Then the image is segmented according to the color information in HSV space, and the independent connected domains of the segmented blocks are solved in combination with the mathematical morphology method. Afterwards, the edge growth method is used to restore the disparity map of each connected domain. Finally, left-right consistency check is applied to improve the accuracy of disparity map. The experimental results show that it is effective to measure the similarity between the matching points of the solid color and the weakly textured area by using the value features of the image. The V-SAD algorithm improves the SAD algorithm''s matching result in regions with weak texture and solid color, and the average mismatch rate decreases by 13.02%.