基于兴趣点和边缘特征的图像检索方法的研究

Image retrieval based on combining interest points and edges for content

提出了多种图像特征相结合的图像检索新方法。该方法把兴趣点和边缘作为图像中用户关注的线索,设计了一种基于兴趣点和边缘特征的图像检索新方法。它综合考虑了图像的全局形状特征、颜色及其空间分布信息,不仅克服了单一特征无法真正表征图像的缺陷,同时保证了检索算法对图像旋转、平移的识别不变性。进行检索时,可根据兴趣点的数目自适应地改变兴趣点和边缘特征的相似性权重,增加了系统的有效性。与同类方法相比较表明,该方法有效提高了图像检索的准确率。

A novel approach using combined features to retrieve images is presented. Interest point and edges are regarded as the vision clues, and a novel approach is designed. Considered the overall features of the image, color and spatial distribution, it not only overcomes the shortage that a single image features can not be truly characterized, but also has a strong robustness to rotation, translation and criterion. Experimental results show this system has good performance.