关键词: 基于内容的图像检索,无线胶囊内镜图像,色调饱和度相关直方图,LBP纹理特征,高斯归一化,特征融合

Abstract: Wireless capsule endoscopy (WCE) is a new technique for detecting the gastrointestinal diseases which has a great market value．This technique will release a mass of image data and lead a big challenge for doctors to review the WCE images．Hence,WCE image retrieval is a fundamental function of the WCE image analysis system．This paper investigates the WCE image retrieval problem by using the similarity information of the WCE images．With the analysis of the property of WCE image,we develop an image retrieval algorithm by fusing the color feature and texture feature．Firstly,the color feature hue-saturation correlation histogram and texture feature LBP (local binary pattern) have been extracted；Then the color and texture features are fused by Gaussian normalization；Finally the index of the WCE image dataset can be built accordingly using the resultant fused feature vector．The experiments show that the proposed method has low computational complexity and a good retrieval result.

Key words: Content-based image retrieval,Wireless capsule endoscopy,Hue-Saturation correlation histogram,LBP,Gaussian normalization,Fused feature