论文部分内容阅读
计算机立体视觉中 ,获取含单目特征的立体图像对中特征的视差分布一直是个难点 .有别于传统的基于特征属性的匹配 ,本文在FACADE(Form And Color And Depth)视觉理论及其神经元动力学方程基础上 ,构造出FA CADE“双侧竞争”双目滤波器 ,它采用不同的竞争策略对表示单目特征的神经元和表示双目特征的神经元分别进行处理 ,从而将立体图像对中处于不同深度上的特征分配到不同的神经元表示平面上 .实验结果表明用神经元动力学方法来获取包含单目特征的立体图像对的视差分布是可行的 .
In computer stereoscopic vision, it is always a difficult point to obtain the parallax distribution of stereoscopic image pairs containing monocular features.Different from the traditional matching based on feature attributes, this paper presents a new algorithm called FACADE (Form And Color And Depth) vision theory and its neurons Based on the dynamic equations, a FA CADE “bilateral competition” binocular filter is constructed, which uses different competitive strategies to process neurons representing monocular features and neurons representing binocular features, respectively, so that the stereoscopic images The features at different depths are assigned to different neuron representation planes.The experimental results show that it is feasible to obtain the parallax distribution of the stereo image pairs containing monocular features by using neuron dynamics.