基于偏态分布混合模型的组织学图像颜色标准化  

Color Standardization of Histological Images Based on Skewed Distribution Mixture Model

作　　者：范小燕 王瑞 FAN Xiaoyan;WANG Rui(Zhengzhou University of Science and Technology,Zhengzhou 450064,China)

机构地区：[1]郑州科技学院,河南郑州450064

出　　处：《现代信息科技》2025年第5期169-173,共5页Modern Information Technology

摘　　要：组织学图像常用的染色方法为苏木精-伊红(H&E)染色法。然而,因染色原材料、储存条件或染色时间等因素的差异,会致使组织学图像的颜色产生变化,进而对计算机辅助组织学图像分析的性能造成影响。因此,在组织学图像分析中,降低图像间的颜色差异并留存图像所包含的组织学信息极为关键。文章引入一种基于偏态分布混合模型的组织学图像颜色标准化方法,用以消除组织学图像之间的颜色变化,并在三个数据集上开展了广泛实验,同时将其与常用的颜色标准化方法进行对比。实验结果显示,该方法在颜色标准化方面具备更优的性能。The commonly used staining method for histological images is Hematoxylin-Eosin(H&E)staining.However,due to the differences in staining raw materials,storage conditions or staining time,the color of histological images can change,which affects the performance of computer-aided histological image analysis.Therefore,in histological image analysis,it is very important to reduce the color difference between images and retain the histological information contained in the images.In this paper,a color standardization method of histological images based on skewed distribution mixture model is introduced to eliminate the color changes between histological images.Extensive experiments are carried out on three data sets,and it is compared with the commonly used color standardization methods.The experimental results show that the method has better performance in color standardization.

关 键 词：偏态分布 颜色标准化 EM算法 非负矩阵分解 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]