融合多特征和相关反馈的图像检索  

Image Retrieval Based On Multi-Feature Fusion and Relevance Feedback

作　　者：陈慧[1] 林智华[1] 黄堃[1] 林雯[1] 林旭云[1] CHEN Hui LIN Zhi-- hua HUANG Kun LIN Wen LIN XU-- yun(College of Electronics and Information Science,Fujian Jiangxia University, Fuzhou 350108, China)

机构地区：[1]福建江夏学院电子信息科学学院,福州350108

出　　处：《德州学院学报》2016年第6期42-46,共5页Journal of Dezhou University

基　　金：福建省教育厅A类科技项目(JA14321)

摘　　要：快速高效的图像检索算法是基于内容的图像检索的重要技术.针对单一特征检索在图像检索中的弊端,充分利用特征互补,提出了一个多特征融合与相关反馈的图像检索算法.首先使用环形直方图法结合图像的空间关系,提取图像的颜色特征,然后使用四叉树分块法并运用空间局部特征增强图像的全局特征,提取图像的形状特征,最后融合图像的颜色特征和形状特征作为综合特征,并根据相关反馈调整参数,实现图像检索.实验结果表明,该算法能提高图像检索的效率,优于使用单一特征的检索方法,尤其适用于基于对象或区域的图像检索.Efficient image retrieval algorithm is based on the key technology of image retrieval. Aiming at the deficiency of single feature retrieval in image retrieval, take full advantage of the complementary char- acteristics, an image retrieval algorithm base on multi--feature fusion and relevance feedback is proposed. Firstly, combined with the spatial relationship of the image, the color feature of image was extracted by u- sing Ring color histogram, then space utilization of local information to enhance the overall image charac- teristics, shape feature of image was extracted by using quadtree block method, and finally integration of color features and shape features as an integrated feature, and in accordance with the relevant feedback to adjust the parameters for image retrieval. The experimental results show that the algorithm can improve the image retrieval efficiency over the use of single feature retrieval method, especially suitable for the ob- ject-based or region--based image retrieval.

关 键 词：图像检索 颜色特征 形状特征 相关反馈 环形颜色直方图 四叉树 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]