基于边缘检测的多特征智能图像检索模型  

A Multi-Feature Intelligent Image Retrieval Model Based on Edge Detection

作　　者：史婷婷[1] 刘卫华[2] 吴明珠[3] 

机构地区：[1]仲恺农业工程学院信息科学与技术学院,广州510225 [2]广东司法警官职业学院网络中心,广州510520 [3]华南师范大学增城学院计算机系,广州511363

出　　处：《华南师范大学学报（自然科学版）》2014年第1期27-31,共5页Journal of South China Normal University(Natural Science Edition)

基　　金：国家自然科学基金项目(31100510);仲恺农业工程学院校级科研基金项目(G3081807)

摘　　要：提出一种智能图像二级检索模型.第一级检索利用改进的Canny算子和颜色向量角分别检测图像边缘,基于db2小波进行图像融合得到二值边缘,并运用图像位错率进行边缘信息比较,将相似度大的图像组成一个新的备选图像库,以缩小图像检索范围.然后利用"HSI颜色不变量模型"改进颜色相关矩阵,加以边缘直方图进行综合特征第二级检索.实验表明该模型能够高效和准确地查找出用户感兴趣的彩色图像,并具有较好的查全率和查准率.A two-stage intelligent image retrieval model is proposed. In the first rank the model detects the edge of original image by the means of new Canny operator and color vector angle respectively, and then it conducts image fusion of the two results that produced by those two detection methods based on db2 wavelet, and finally bit error rate is used to compare the similarity of edge information in order to narrow the retrieval range, which turned out a new alternative image library. In the second rank HSI color invariant model is used to improve the color correlation matrix for multi-feature image retrieval with edge histogram. Experimental results show that our model is efficient and accurate in the user-interested images retrieval with better recall and precision.

关 键 词：智能图像检索 边缘检测 CANNY算子 颜色向量角 边缘直方图 颜色不变量 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]