论文部分内容阅读
首先在图像间提取反映图像内在几何布局和形状属性的局部自相似(LSS)描述子,之后结合LSS和归一化相关系数(NCC)构建了一种形状相似性测度——LSCC,并采用模板匹配的策略识别同名点。对比NCC和互信息,LSCC在顾及计算效率的同时获得了更好的匹配正确率。另外,根据遥感图像的特点,设计了一种基于LSCC的自动配准方法。试验结果表明,该方法能够较好地抵抗图像间的非线性灰度差异,并获得可靠的配准精度。
Firstly, a local self-similarity (LSS) descriptor reflecting the intrinsic geometric layout and shape attributes of the image is extracted from the images, and then a shape similarity measure - LSCC is constructed based on the LSS and the normalized correlation coefficient (NCC) Matching strategy to identify the same name point. Compared with NCC and mutual information, LSCC achieves better matching accuracy while considering computational efficiency. In addition, according to the characteristics of remote sensing images, an automatic registration method based on LSCC is designed. Experimental results show that this method can resist nonlinear grayscale difference between images and obtain reliable registration accuracy.