基于内容的图像检索-感兴趣区域的提取

Content-based Image Retrieval—Extracting Region of Interest

基于内容的图像检索是当前研究的热点和难点,提取图像的感兴趣区域,能够提高检索的准确性.该文提出了一种基于特征点的感兴趣区域提取方法,首先基于多曲率多项式提取图像的特征点,然后根据一定的规则提取感兴趣区域,实验证明,使用这种感兴趣区域提取方法,能较大地提高检索的查准率.

Content-based image retrieval is the hotspot and aporia of current research,extracting the region of interest can achieve good performance. A novel method of extracting the region of interest is presented based on salient points. The salient points are firstly detected by multi-scale curvature polynomial,then extracts the region of interest according to some rules. Experimental results show that the method can effectively describe the region of interest and achieve good performance.