一种基于树结构的立体图像特征提取算法

针对立体图像匹配中的特征检测问题，提出了一种基于树结构的立体图像中边缘点、直线段、交点和凸多边形区域等特征的检测方案。该方法首先基于改进的Canny算子实现立体图像边缘的检测；基于立体图像的边缘梯度信息，文中提出一种分层并行式迭代式链码跟踪方法实现图像中直线段的检测：为提高检测速度并有利于凸多边形区域特征的检测，基于凸多边形区域的几何定义，通过提取直线段交点并采用树结构实现对已检测的相关直线段信息的遍历处理，最终得到树结构表示的凸多边形区域特征。实验证明，所提方法在结构化环境具有很好的检测精度和实时性。

Efficient stereo-image feature detection algorithm based on tree structure

Aiming at feature detection for stereo matching,research on the feature extraction of stereo image such as edge,line segment,corner and polygonal region was performed. Firstly,an improved Canny edge detection algorithm was represented to obtain thin and robust edges. Secondly,an iterative chain code tracing method with parallel hierarchy was introduced for line segment detection based on detected edge gradient of image. According to the geometrical definition,corners refered as the endpoints where two or more line segments meet were employed and polygonal regions composing tree nodes with line segments and corners were detected through traversing relevant line segments and corners in tree structure. Experiments on practical stereo images in structured environment have demonstrated the significant performance of proposed approach.