基于改进自适应权重的三边滤波立体匹配算法

针对基于双边滤波器（BF）的自适应权重（ASW）方法不能有效解决由视差不同但颜色相似的像素引起的模糊匹配问题，引入了一种新的基于三边滤波器（TF）的ASW方法，通过局部能量模型计算相邻像素之间的边界强度来提高匹配精度。为了提高匹配速度，将TF算法递归实现，把普通局部立体匹配算法的复杂度从O(NWD)]降低为O(N)]。在Middlebury基准测试集上进行实验并与其他局部立体匹配算法进行比较，RTF算法的平均误匹配率为4.91%，匹配精度高于同类型双目立体匹配算法，平均匹配速度达到258 ms，满足了双目立体匹配实时性的需求。

Three-Edge Filtering Stereo Matching Algorithm Based on Improved Adaptive Support Weight

The Adaptive Support Weighting（ASW） method based on Bilateral Filter（BF） can not effectively solve the fuzzy matching problem caused by pixels with different parallax but similar colors, a new Trilateral Filter（TF） based ASW method is proposed. The local energy model is used to calculate the boundary strength between adjacent pixels to improve the matching precision. In order to improve the matching speed, the TF algorithm is recursively implemented, and the complexity of the ordinary local stereo matching algorithm is reduced from O(NWD)] to O(N)]. Experiments are carried out on the Middlebury benchmark test set. Compared with other local stereo matching algorithms, the average mismatch rate of the RTF algorithm is 4.91%, higher than the other binocular stereo matching algorithm. The average matching speed reaches 258 ms, which satisfies stereo matching needs for real-time.