摘要
乳腺癌诊断的图像处理过程主要包括以下三个步骤:感兴趣区域(ROI)提取、图像增强和特征提取.由于传统的图像增强方法是应用在整个ROI上的,因此ROI中不相关或无用信息的增强会转化为劣质特征.为了解决这一问题,提出了基于信息熵的图像局部增强策略.该策略对每幅乳腺图像的ROI进行局部分割,选择熵值最大的区域块.通过多轮的图像增强策略进一步改进优胜块,并嵌入到原始ROI中.在此过程中,将由熵权法计算结果值最大的一组特征来表示这幅图像.实验结果表明,该方法提取的特征在分类精度和AUC指标方面优于原始图像、全局增强图像和随机局部增强图像的特征.
The procedure of image processing for breast cancer diagnosis mainly consists of three steps:the region of interest(ROI)extraction,image enhancement and feature extraction.Since the conventional image enhancement is implemented on the whole area of the ROI,the irrelevant or useless information in ROI get enhanced and transformed to the inferior features.In this paper,an image local enhancement strategy based on information entropy is studied for addressing such problem.By the proposed strategy,the ROI of each mammographic image is segmented to select the block which contains the highest value of the entropy.The winner block will be further improved by a multi-round image enhancement strategy and embedded into the original ROI.In doing so,each image will be represented by a set of features of the maximum value calculated by the entropy weight method.Experimental results show that the features extracted by this strategy are superior to the features of original image,global enhanced image and random local enhanced image in terms of classification accuracy and AUC.
作者
付其林
邓安生
曲衍鹏
FU Qi-lin;DENG An-sheng;QU Yan-peng(School of Information Science and Technology,Dalian Maritime University,Dalian 116026,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2020年第4期820-824,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61502068)资助
大连市青年科技之星项目(2018RQ70)资助。
关键词
图像处理
特征提取
信息熵
图像局部增强
image processing
feature extraction
information entropy
image local enhancement
