摘要：

提出一种基于同构径向基函数(RBF)网络的视差图像分治重建方法，通过可调域值和边缘检测完成视差图像的区域分割，在每个区域内采用RBF神经网络进行重建，将各个区域进行拼接，得到最终的重建结果。在不同区域运用RBF进行重建时根据各区域结构特征的繁简用不同分辨率的数据进行训练。实验结果表明，该方法可以获得高质量的重建结果。

关键词: 视差图像, 立体视觉, 径向基函数网络, 三维重建

Abstract:

A partition reconstruction method for disparity image based on homogeneous RBF network is presented. Disparity image is divided into some regions by reconcilable threshold and edge detection. Reconstruction based on RBF neural network is carried out in every region. All regions are combined and reconstruction result comes into being. When reconstruction based on RBF is executed in different region, data which have different resolving power are used in training according to structure of different region. Experimental results show that good reconstruction can be attained by the using method proposed.

Key words: disparity image, stereo vision, RBF network, 3D reconstruction

中图分类号: 

• TP391