基于NALTDP的肝脏CT图像分类

Liver CT Image Classification Based on Neighborhood Average Local Ternary Direction Pattern

针对肝功能正常和肝功能异常CT图像之间差异性不明显的问题,提出一种基于邻域变化的特征提取模式,称为邻域均值局部三值方向模式(NALTDP),用于肝脏CT图像的分类。先对肝脏CT图像预处理,提取大小相同的正常和异常图像;再利用中心像素在相邻八个方向上邻域变化的相关性计算邻域均值,并对邻域均值进行三值模式、方向模式和三值方向模式的编码;最后统计直方图并采用支持向量机分类识别。NALTDP算子通过提取各个方向的邻域变化信息和方向信息,表达更多层次的局部纹理信息,放大图像间的差异性。实验表明,在不同参数情况下,与已有的几种方法相比,NALTDP的分类识别率明显更高,所提出的分类方法比现有方法具有更好的性能。

To solve the problem that the difference between normal and abnormal liver CT images is not obvious,a feature extraction mode based on neighborhood change,called neighborhood average local ternary direction pattern(NALTDP),is proposed to classify liver CT images.Firstly,liver CT images were preprocessed to extract normal and abnormal images of the same size.Then,the neighborhood mean was calculated by using the correlation of the neighborhood changes of the center pixel in the adjacent eight directions,and the neighborhood mean was encoded by the three-valued mode,the direction mode and the three-valued direction mode.Finally,the histogram was counted and classified by support vector machine.NALTDP operator can extract neighborhood change information and direction information in each direction to express local texture information of more levels and enlarge the difference between images.Experimental results show that under different parameters,the classification recognition rate of NALTDP is significantly higher than that of the existing methods,and the proposed classification method has better performance than the existing methods.