基于NSCT边缘直方图的草图检索  

Edge histogram for sketch based image retrieval using nonsubsampled contourlet edge detection

作　　者：陈曦[1] 肖建[1] 

机构地区：[1]长沙理工大学计算机与通信工程学院,长沙410114

出　　处：《计算机工程与应用》2015年第22期172-175,186,共5页Computer Engineering and Applications

基　　金：湖南省自然科学基金项目(No.11JJ3069)

摘　　要：基于草图的图像检索,存在着大量的应用空间。针对查询草图和数据库图像的差异,提出了一种利用非降采样Cotourlet(NSCT)变换和边缘直方图的草图检索算法。利用NSCT变换分解图像,对所有子带进行边缘检测,提取子带的边缘直方图用来作为描述图像和草图的特征。将该方法与边缘直方图比较实验表明,其检索效果更准确。There is a wide range applications of Sketch Based Image Retrieval. Considering the differences of sketches and images, it is to be proposed in this paper that a new approach of Sketch Based Image Retrieval through using Non Subsampled Contourlet（NSCT）transform and Edge Histogram Descriptor. By using NSCT transform decomposition of image to do edge detection for all subbands, and extract the edge histogram of the subbands that is used as describing the features of images and sketches. After comparing the method to the edge histogram descriptor, the experimental results show that the retrieval results are more accurate.

关 键 词：草图检索 非降采样Contourlet 边缘直方图 边缘检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]