摘要： 提出了一种基于核密度估计相似性测度的立体匹配方法。将匹配基元中对应像素点的颜色差值在差值特征空间中进行核密度估计，并以此作为相似性测度。在此基础上建立全局能量函数，并采用改进的置信度传播算法进行视差图的求解。实验结果证明该相似性测度较传统方法有明显的改进。

关键词: 立体匹配, 核密度估计, 差值样本空间, 置信度传播

Abstract: A new stereo matching method based on Kernel Density Estimation(KDE) similarity is proposed. KDE of the color difference value between the corresponding matching elements is calculated in difference feature space and selected as the similarity measure. Based on this similarity measure, a global energy function is constructed and the final disparity map is obtained by the improved belief propagation algorithm. Experimental results demonstrate the improvement of the measure compared with the traditional methods.

Key words: stereo matching, Kernel Density Estimation(KDE), difference sample space, belief propagation

中图分类号: 

• TP391