本文讨论一种基于亮度对立体图对左右图象的视差进行估计的方法。因为最简单的基于亮度的方法 ,如块匹配法 ,在尺寸块较小时对噪声比较敏感 ,所以近来人们提出了一种基于非线性扩散的比传统块匹配法有更好的噪声抑制性能的改进块匹配法。本文将探讨这种新方法在由左右摄像机增益差而导致立体图对亮度和反差不同时的情况。首先我们指出改进块匹配法与传统方法相比对上述亮度和反差方面的差别更不敏感。其次我们介绍一种基于立体图对统计特性的称为增益控制的方法以改进视差估计算法的性能。 This paper discusses a method of estimating the disparity of the left and right images based on the brightness and stereo image. Since the simplest brightness-based methods, such as block matching, are relatively noise-sensitive when the size of the block is small, recent improvements based on nonlinear diffusion have been shown to have better noise suppression performance than traditional block matching Block matching method. This article will explore the new method in the left and right camera gain difference caused by the perspective of the brightness and contrast is not the case. First of all, we point out that the improved block matching method is less sensitive to the differences in brightness and contrast as compared with the traditional methods. Second, we introduce a method called gain control based on the statistical properties of stereo pairs to improve the performance of the parallax estimation algorithm.