基于边缘融合的多图像直线提取算法

Line Extraction in Multiple Images Based on Fusion Edges

该文提出了一种新的基于边缘融合的多图像直线提取算法 ,该算法首先对所有多图像分别提取边缘 ,然后通过将多图像对应的梯度方向相似的、相邻近的边缘编组 ,以形成融合的直线支持区域 ,最后基于最小二乘的方法从直线支持区域拟合出直线段 ,并提取直线属性 .该文提出的算法把边缘的相位信息作为融合要素 ,能够从复杂的图像中提取出相当低对比度的直线 ,并且通过分析多图像边缘之间的特性 ,提出相位压缩和融合编组方案 ,不仅可以充分融合互补信息还有效降低了算法复杂度 .真实的遥感图像直线提取结果表明了该文算法的有效性 .

Linear feature is one of the most essential features that used in computer vision and image understanding. Effective approaches of line extraction have been paid more attentions so far. The techniques of image fusion can effectively extract useful information from a set of input images. A new line extraction algorithm in multiple images based on fusion edges is proposed, It consists of three main aspects: Firstly, edges are extracted from every image. Then, the adjacent edge points from all images are grouped into combined line-support regions based on similarity of gradient orientation. Finally, the regions are fitted and the fused lines are generated. This method uses the gradient orientation of multiple images as the fusion element. More complementary information is used, and more effective lines can be extracted. The validity of the method is proved by experiments for real remote sensing images.