多特征分层融合医疗设备图像检索方法  

Medical Image Retrieval Method Based on Multi Feature Hierarchical Fusion

作　　者：单强[1] 孙晓明[2] 

机构地区：[1]首都医科大学附属北京朝阳医院,北京100020 [2]哈尔滨理工大学测控技术与仪器黑龙江省高校重点实验室,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2017年第2期135-139,共5页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(61401126);黑龙江省自然科学基金(QC2015083)

摘　　要：针对解决医疗设备图像的检索问题,提出了一种基于多特征分层融合的图像检索方法。此方法采用颜色矩计算图像的颜色特征,采用LBP算子计算图像的纹理特征,采用Fourier描述子计算图像的形状特征,通过分层加权融合的方法将三类特征融合在一个相似性测度之内,最终形成检索判断的依据。根据窥镜图像和CT图像展开实验研究,其结果证实了所提出方法的有效性。随着查全率要求的不断提升,所提出方法的查准率明显高于三种单一特征的检索方法,适合于医疗图像的检索。Objective： In order to solve the problem of medical equipment image retrieval, an image retrieval method based on multi feature layer fusion is proposed. Process： This method using color moments to calculate the characteristics of color image, the LBP Operator is used to calculate the texture features of the image, using Fourier descriptors calculated image of the characters, the stratified weighted fusion method of the characteristics of three types of fusion within a similarity measure, and ultimately the formation of retrieval based on. Results and conclusions： According to the experimental study of the endoscopic image and CT image, the experimental results confirm the validity of the proposed method. With continuous improvement of the precision requirements of the proposed method in precision ratio was significantly higher than that of two single feature retrieval method for medical image retrieval.

关 键 词：图像检索 多特征 分层融合 医疗设备 

分 类 号：N393[自然科学总论]