考虑空域分布多尺度信息熵的SAR图像模板匹配

图像的熵和多尺度熵仅考虑像素灰度分布而无视像素在空域分布的情况,基于此的图像匹配容易受噪声的影响而导致误配。为解决此问题,给出了一种空域分布多尺度信息熵(SDMSE),将图像像素在空域的分布与灰度空间分布结合起来,对不同的行或列求多尺度信息熵。在合成孔径雷达(SAR)图像匹配时,对输入图像和基准子图(基准图中和输入图尺寸一样的子图)求SDMSE矩阵,并通过求两矩阵的相似性来度量匹配程度,相似性最大的位置对应匹配点。仿真结果表明,所提匹配算法相比基于熵和多尺度熵的SAR匹配算法有更优异的噪声适应性,匹配误差更小,但计算耗时较多。在如何减少计算时间方面也做了尝试,实验表明尺度个数减少可以大幅减少计算时间而抗噪声性能并没有明显降低。

SAR image template matching based on spatial distribution multi-scale entropy

When calculating the entropy and multi-scale entropy of image,only intensity distribution of pixels is considered,but the pixel's position information is ignored.Image matching based on entropy and multi-scale entropy is vulnerable to noise so as to lead to mismatches.In order to solve this problem,spatial distribution multi-scale entropy(SDMSE) is proposed.The spatial distribution and intensity distribution of the image pixels are combined.Multistage entropy of each row or column is calculated.In the synthetic aperture radar(SAR) image matching,both the input image and the reference sub-graph are calculated for SDMSE matrix.Correlation coefficients of two matrices are calculated,the position of the maximum of correlation coefficients corresponds to matching point.Experimental results show that the proposed algorithm has better noise adaptability than the algorithms based on multi-scale image entropy and entropy,and its reliability is better.But the calculation is more time-consuming.How to reduce computation time is tried and experiments show that reducing scale number can substantially reduce the computational time but the anti-noise performance does not significantly reduce.