编码选择哈希算法解决大规模图像检索问题（英文）

发布时间：2018-12-11 14:15

【摘要】：哈希算法已被广泛应用于解决大规模图像检索的问题.在已有的哈希算法中,无监督哈希算法因为不需要数据库中图片的语义信息而被广泛应用.平移不变核局部敏感哈希(SKLSH)算法就是一种较为代表性的无监督哈希算法.该算法随机的产生哈希函数,并没有考虑所产生的哈希函数的具体检索效果.因此,SKLSH算法可能产生一些检索效果表现较差的哈希函数.在本文中,提出了编码选择哈希算法(BSH).BSH算法根据SKLSH算法产生的哈希函数的具体检索效果来进行挑选.挑选的标准主要根据哈希函数在3个方面的表现:相似性符合度,信息包含量,和编码独立性.然后,BSH算法还使用了一种基于贪心的选择方法来找到哈希函数的最优组合.BSH算法和其他代表性的哈希算法在两个真实图像库上进行了检索效果的对比实验.实验结果表明,相比于最初的SKLSH算法和其他哈希算法,BSH算法在检索准确度上有着明显的提高.
[Abstract]:Hash algorithm has been widely used to solve the problem of large-scale image retrieval. Among the existing hash algorithms, the unsupervised hash algorithm is widely used because it does not require semantic information of images in the database. The translation-invariant kernel locally sensitive hash (SKLSH) algorithm is a typical unsupervised hash algorithm. The algorithm generates hash function randomly without considering the specific retrieval effect of the generated hash function. Therefore, the SKLSH algorithm may produce some hash functions with poor retrieval performance. In this paper, a coding selective hash algorithm (BSH). BSH) is proposed, which is selected according to the specific retrieval effect of the hash function generated by the SKLSH algorithm. The selection criteria are based on the performance of the hash function in three aspects: similarity coincidence, information inclusion, and coding independence. Then, the BSH algorithm also uses a greedy selection method to find the optimal combination of hash functions. The BSH algorithm and other representative hash algorithms are compared in two real image databases. Experimental results show that, compared with the original SKLSH algorithm and other hash algorithms, the BSH algorithm has a significant improvement in retrieval accuracy.
【作者单位】：华南理工大学计算机科学与工程学院;
【基金】：Supported by National Natural Science Foundation of China(61272201,61572201) National Undergraduate Innovative and Entrepreneurial Training Program(201510561072)
【分类号】：TP391.41

【相似文献】