To shorten the convergence time, the optimal search problem of disparity map is converted to an iterative convergence process of bi-valued neural networks. 为加快收敛速度,该算法将视差图的最优搜索问题转换为二值神经网络的迭代收敛过程。
Finally, the disparity of both matched and occluded points are obtained through finding the optimal path using dynamic programming, and the dense disparity map is obtained. 最后通过动态规划寻找最优路径来获得匹配点和遮挡点的视差,从而得到稠密视差图。
The information of both matched and occluded points are obtained through finding the optimal path using dynamic programming, and occlusion is detected and handled by disparity consistency testing and the occluded pixels filling, and then the dense disparity map is obtained. 该方法采用动态规划寻找最优路径来获得匹配点的视差和遮挡点信息,并利用视差一致性检验和遮挡像素视差填充等过程来实现对匹配过程中遮挡的检测和处理,从而得到稠密视差图。
The method processes a global optimal motion estimation in overlapped window region, and use the hierarchical structure to overcome the disadvantages of fixed size block matching, so have a correct disparity vector, and produce a smoother disparity map. 由于该方法在交叠窗口区域进行全局最优估计,并且采用多分辨率的结构克服固定块匹配的缺陷,得到较准确的视差矢量,并且产生光滑的视差图。