摘要
提出一种新的基于ROI(regions of interesting)的多分辨率网格框架的图像检索方法。先利用网格对图像进行分块,并利用小波变换对分块进行多分辨率分析,以提取图像各分块的颜色和纹理特征信息,然后根据用户指定的感兴趣的区域得到查询子图,最后使用一种融合子图(integrated sub-images matching,ISM)的匹配方法进行图像检索。通过对比实验,证明该方法能够有效地得到降低图像检索的语义鸿沟,并取得很好的检索效果。
A novel image retrieval method of multi-resolution grid framework was proposed based on the regions of interesting （ROI）. First the image was segmented by grid, and then the image blocks were multi-resolution analyzed by wavelet transform to extract color and texture features. Secondly, the system obtains sub-images by user specified blocks which contain regions of interesting information. At last, an integrated sub-images matching （ISM） scheme was used to compute image similarity. The contrast experimental results demonstrate that this method can reduce the gap between low-level visual feature and high-level semantic information and obtain good retrieval effect.
出处
《重庆邮电大学学报（自然科学版）》
北大核心
2009年第5期686-689,共4页
Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)
