基于视觉显著图的彩色图像检索  

Color Image Retrieval Based on Visual Saliency Map

作　　者：胡胜雄[1] 尚赵伟[1] 张太平[1] 杨维斌[1] 

机构地区：[1]重庆大学计算机学院,重庆400030

出　　处：《计算机工程》2012年第8期189-191,共3页Computer Engineering

基　　金：国家自然科学基金资助面上项目(61003120);重庆市自然科学基金资助项目(CSTC2010BB2217);中央高校基本科研业务费基金资助项目(CDJRC10180009)

摘　　要：为提高图像检索效率,提出一种基于视觉显著图的彩色图像检索方法。利用视觉显著图去除原图中与检索任务无关的背景信息,保留用户感兴趣的图像区域信息,采用小波域的BDIP-BVLC方法提取图像特征,并引入二次查询度量策略进行距离度量。实验结果表明,与基于显著性加权的检索方法相比,该方法的平均查准率较高。In order to improve efficiency of image retrieval, a retrieval method based on visual saliency map is proposed in this paper. The background information unrelated to retrieval task in original image is removed according to the visual saliency map, and remains the region of interest. The BDIP-BVLC moments in the wavelet domain is used to extract the image features. The secondary inquires measure strategy is adopted as the distance function. Experimental result shows that the proposed method outperforms the method based on saliency weighted. It has higher retrieval accuracy.

关 键 词：感兴趣区域 视觉显著图 小波 BDIP—BVLC特征 二次查询度量 图像检索 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]