Research Initiation: Stereo Matching Using Topographic Structures

解决立体匹配问题后，可以通过简单的三角变换确定立体图像中可见表面的三维位置。现有的立体匹配方法大多涉及边缘等底层图像特征的匹配。本文提出了一种基于高级图像特征匹配的改进立体匹配方法。我们将探索提取高级图像特征的新技术，目的是提取图像中灰度强度变化的丰富符号表示。将开发用于确定在立体图像中提取的图像特征之间的对应关系的方法。目标是开发用于生成立体图像的全视差图的算法。

Three-dimensional positions of the visible surfaces in stereo images can be determined by simple trigonometric transformation once the stereo matching problem is solved. Most existing stereo matching methods involve matching low-level image features such as edges. This research describes an improved stereo matching procedure based on matching high-level image features. New techniques for extracting high-level image features will be explored with the aim of extracting a rich symbolic representation of the gray intensity changes in an image. Methods will be developed for determining correspondences between the extracted image features in the stereo images. The goal is to develop algorithms for generating full disparity maps for stereo images.