摘要
随着互联网的迅速发展,产生了大量的近重复视频。文章提出了一种改进的哈希算法提高近重复视频的检索准确性,根据语义哈希对图像检索的原理,对算法中的邻接矩阵进行改进。邻接矩阵表示KNN图中样本间的邻接关系,文中不再使用0和1两个值表示样本间的邻接关系,而是引入高斯核函数来表示,提高了模型的检索精度。实验结果表明所提出的方法具有更高的检索精度。
With the development of Internet, a large number of near-duplicate videos are produced on- line each day. In this paper, an improved hashing algorithm is proposed to improve the accuracy of near-duplicate video retrieval. According to the theory of semantic hashing based image retrieval, the adjacency matrix is improved. Adjacency matrix is a representation of the sample＇s adjacency of K-nearest neighbor（KNN） graph. The adjacency relationship is presented by Gaussian kernel function instead of 0 or 1, thus improving the accuracy of retrieval. The experimental results show that the proposed method has higher retrieval accuracy.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2016年第1期67-72,共6页
Journal of Hefei University of Technology：Natural Science
基金
安徽省科技攻关计划资助项目(1301b)
关键词
近重复视频检索
哈希算法
邻接矩阵
高斯核函数
KNN图
near-duplicate video retrieval
hashing algorithm
adjacency matrix
Gaussian kernel function
K-nearest neighbor（KNN） graph
