基于颜色特征相似度判别的肿瘤图像检索研究  

Research on tumor image retrieval system based on relevancy discriminant by color feature

作　　者：李广丽[1] 张红斌[2] 

机构地区：[1]华东交通大学信息工程学院,江西南昌330013 [2]华东交通大学软件学院,江西南昌330013

出　　处：《计算机工程与设计》2012年第11期4272-4277,共6页Computer Engineering and Design

基　　金：2011年度教育部人文社科基金项目(11YJC870012);2011年度江西省教育厅青年基金项目(GJJ11106);2011年度华东交通大学校立基金项目(11XX02);2012年江西省科技专项基金项目(20121BBG70050)

摘　　要：给出医学图像检索的工作流程,分析并选择颜色特征作为医学图像检索的重要依据。经过颜色空间选择、颜色空间转换、颜色量化和颜色特征提取完成医学图像特征抽取,设计了基于欧式距离、加权欧式距离和直方图相交的相似度判别算法。实验结果表明,检索单幅图像的处理时间不超过0.432s;从平均值角度考虑,正常图像的召回率优于其它类别图像;正常图像检索的准确率高于癌变图像检索的准确率。The working flow of a medical image retrieval system is put forward firstly, then some image features in a tumor image are analyzed and color features are chosen as the basic design principle to make correct image retrieval. Some pretreatment steps must is done firstly such as color space selection, color space transformation, color quantification and color features selection. With the aid of pretreatment, three relevancy discriminant algorithms such as Euclidean Distance, Improved Euclidian Distance and Cross Color Histogram Statistics are in turn designed and realized to calculate the ＂relevancy distance＂. Finally, the performances of each algorithm are gotten and some conclusions are also deduced. The time processing of each image is no more than 0. 432s. While many recall rates are gotten that the recall rate of the normal images is superior to any other kinds of images. Retrieval accuracy of normal image is higher than that of cancer image.

关 键 词：肿瘤 图像检索 颜色空间 相似度判别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]