A novel method using KDE and Graph Cut in Stereo Matching

This paper presents a new stereo matching method based on Kernel Density Estimation (KDE) and Graph Cut. KDE similarity is applied to the image´s color disparity difference sample space and a measure method of the difference of the corresponding matching elements is obtained. Based on the KDE similarity, the data penalty of the Global Energy Function of Stereo Matching problem can be constructed. And final disparity map is obtained by the Graph Cut algorithm. Experimental results of the proposed algorithm demonstrate our method comparing with the traditional methods.