KAZE算法在图像检索中的应用

摘要

An image retrieval algorithm combining KAZE algorithm with the visual dictionary model is proposed.In this algorithm,the KAZE algorithm is used to extract feature points,to construct feature descriptors,and to pre-match;the random sampling consistency algorithm is used to remove mismatched points and error points;and the K-Means algorithm is also used to characterize descriptors clustering.Then the visual dictionary model is established.The image feature descriptor is then mapped to the visual dictionary model,and the retrieval result is obtained.Experimental results show that the algorithm can retrieve the same or similar image,and the average precision of the image retrieval is 55 ％.%给出一种KAZE算法与视觉词典库模型相结合的图像检索算法.利用KAZE算法提取特征点,构造特征描述子,进行预匹配;采用随机抽样一致性算法剔除误匹配点和错误点;通过K均值聚类算法对特征描述子进行聚类,建立视觉词典库模型.将图像特征描述子映射到视觉词典库模型,得到检索结果.实验结果表明,该算法可以检索相同或相似图像,图像检索的平均查准率为55％.