图像的变化通常没有先验信息，或者即便有也很难用参数模型描述它，当两者都知道时，我们可以得到分布的似然比，构造比检验。-Image changes often do not have a priori information, or even if there is difficult to describe it with the parameter model, when both are known, we can get the distribution of the likelihood ratio to construct ratio test.