中文摘要：

针对光照不均导致传统区域增长算法匹配区域小、误差大等问题,提出一种基于改进Census变换的双重约束匹配算法。用变换窗口灰度均值与局部纹理反差值之和代替中心像素灰度,提高了变换结果的抗噪性和不同子块间的区分度;采用高斯加权变换窗口提高离中心点近的像素点的权重,有效减小了边界不连续区域对匹配的影响。对Middlebury数据库里的图像添加仿真光照,并用其提供的真实匹配值计算验证算法的匹配精度。实验表明较传统算法有更高的匹配精度,且有效匹配点数明显增多。

英文摘要：

Illumination variation results in the problem that the matching area was too small and the matching error was great in the traditional regional growth algorithm. So this paper put forward a matching algorithm of double restraints based on improved Census transform. The sum of gray average and mean difference of relative size replaced the center pixel＇s gray value in transform window. It improved the noise resistance of transformation results and the difference between different blocks. Then it introduced the Gauss template into the Census transformation to improve the weight of pixels closed to center pixel,which effectively reduced the influence of discontinuous area on the matching. This paper added the simulation lighting to the images of Middlebury and used the real matching value to calculate the matching precision. The experiments show that comparing with traditional algorithm,the algorithm can get a higher matching accuracy and more effective matching points.