一种拉近两幅图像相似度的算法

A computation method of improving two images similarity

在图像处理领域中任意两幅图像都有一定的相似度,它往往是用图像的信噪比来衡量的,但是实验结果表明存在代表同一内容的两幅图像的信噪比却低于代表不同内容的两幅图像的信噪比的情况,这是不合理的,文中通过统计分析待求信噪比的两幅图像的有关特性,自动生成一个系数,对结果图像的每一个像素都乘以该系数,可以提高两幅图像的相似度,且绝对不会降低任意两幅的相似度,实验结果和理论分析说明了文中方法的有效性。

It ’ s the same that any two images have a certain degree ol similarity in the field ol image processing. The similarity of two images is often measured by signal- to-noise ratio （ SNR）. However it is unreasonable that there exists such a phenomenon that the SNR of two images which have the same content is lower than that of two images which have different content. In this article,the characteristics of two images are analyzed and a coefficient is automatically generated. If every pixel of the resultant image is multiplied by the coefficient,the similarity of two images is improved and is never reduced. The validity of this method is proved by the results of the experiments and theory analysis.