【摘要】 针对传统检索方法的不足,为了提高图像的检索效果,提出一种兴趣点匹配的图像检索新方法。首先采用Harris算子检测出图像的兴趣点,然后以兴趣点为中心,提取其邻域特征构建图像检索的特征向量,并计算图像特征向量之间的相似性进行兴趣点匹配和图像检索,最后采用Corel图像库进行检索性能的仿真测试。结果证明,相对其它图像检索方法,本文方法的查准率和查全率分别大约提高了6%和4%,可以快速、准确检索用户实际感兴趣的图像。更多还原

【Abstract】 In view of the deficiency of traditional retrieval methods,a new method of image retrieval based on interest points matching is proposed to improve the image retrieval performance. First the new method uses Harris algorithm to detect interest points of image,then it treats points of interest as the center,and extract the neighborhood feature for constructing characteristics of image retrieval to calculate the similarity between the image feature vectors,and image retrieval. Finally use the Corel image database to fultill retrieval performance simulation test. The results show that,comparing with other image retrieval algorithms,the average retrieval accuracy and recall ratio of this method get better to 6% and 4%,which can quickly and accurately retrieve the actual image of the user’s interest.更多还原