Signal Processing Laboratory, Swiss Federal Institute of Technology, Lausanne, Switzerland; Biomedical Engineering and Instrumentation Branch, National Institutes of Health, Bethes.
IEEE Trans Pattern Anal Mach Intell. 1986 Jan;8(1):118-25. doi: 10.1109/tpami.1986.4767760.
The sum and difference of two random variables with same variances are decorrelated and define the principal axes of their associated joint probability function. Therefore, sum and difference histograms are introduced as an alternative to the usual co-occurrence matrices used for texture analysis. Two maximum likelihood texture classifiers are presented depending on the type of object used for texture characterization (sum and difference histograms or some associated global measures). Experimental results indicate that sum and difference histograms used conjointly are nearly as powerful as cooccurrence matrices for texture discrimination. The advantage of the proposed texture analysis method over the conventional spatial gray level dependence method is the decrease in computation time and memory storage.
两个具有相同方差的随机变量的和与差是不相关的，并且定义了它们相关联的联合概率函数的主轴。因此，和与差直方图被引入作为通常用于纹理分析的共生矩阵的替代方法。根据用于纹理特征化的对象类型（和与差直方图或一些相关的全局度量），提出了两种最大似然纹理分类器。实验结果表明，联合使用和与差直方图对于纹理判别几乎与共生矩阵一样有效。与传统的空间灰度依赖方法相比，所提出的纹理分析方法的优点在于计算时间和存储内存的减少。
