摘要
哈希方法因快速及低内存的特点广泛应用于大规模图像检索中,但在哈希函数构造过程中对数据稀疏性缺乏研究。为此,提出一种无监督稀疏自编码的图像哈希算法。在哈希函数的学习过程中加入稀疏构造过程和自动编码器,利用稀疏自编码的KL差异对哈希码进行稀疏约束,以增强局部保持映射过程中的判别性。在CIFAR-10数据集和YouTube Faces数据集上进行实验,结果表明,该算法平均准确率优于DH算法。
The hash method is widely used in large-scale image retrieval due to its fast and low memory characteristics,but it lacks research on data sparsity in the construction of hash functions.To this end,an unsupervised sparse self-encoding image hash algorithm is proposed.In the learning process of the hash function,a sparse construction process and an automatic encoder are added,and the hash code is sparsely constrained by the Kullback-Leibler(KL) divergence of the sparse-auto encoder to enhance the discriminability in the local preservation mapping process.Experiments on the CIFAR-10 datasets and YouTube Faces datasets show that the average accuracy of the algorithm is better than the DH algorithm.
引文
[1] ZHENG Zhaohui,WU Xiaoyun,SRIHARI R,et al.Feature selection for text categorization on imbalanced data[J].ACM SIGKDD Explorations Newsletter,2004,6(1):80-89.
[2] 杨定中,陈心浩.基于投影残差量化哈希的近似最近邻搜索[J].计算机工程,2015,41(12):161-165,170.
[3] 柯圣财,赵永威,李弼程,等.基于卷积神经网络和监督核哈希的图像检索方法[J].电子学报,2017,45(1):157-163.
[4] LIU Wei,MU Cun,KUMAR S,et al.Discrete graph hashing[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems.Cambridge,USA:MIT Press,2014:3419-3427.
[5] 赵文星.变系数部分线性模型的统计推断[D].南京:南京信息工程大学,2015.
[6] XU Jun,XIANG Lei,LIU Qinshan,et al.Stacked sparse autoencoder for nuclei detection on breast cancer histopathology images[J].IEEE Transactions on Medical Imaging,2015,35(1):119-130.
[7] LI Hongmin,LIU Hanchao,JI Xiangyang,et al.CIFAR10-DVS:an event-stream dataset for object classification [EB/OL].[2018-02-05].https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5447775/.
[8] WOLF L,HASSNER T,MAOZ I.Face recognition in unconstrained videos with matched background similarity[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Computer Society,2011:529-534.
[9] JIANG Qingyuan,LI Wujun.Scalable graph hashing with feature transformation[C]//Proceedings of the 24th International Conference on Artificial Intelligence.[S.l.]:AAAI Press,2015:2248-2254.
[10] LIONG V E,LU Jiwen,WANG Gang,et al.Deep hashing for compact binary codes learning[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2015:2475-2483.
[11] MCJUNKIN M C.Precision and recall in title keyword searches[J].Information Technology and Libraries,1995,14(3):161-171.
[12] ABADI M,BARHAM P,CHEN Jianmin,et al.TensorFlow:a system for large-scale machine learning[C]//Proceedings of the 12th USENIX conference on Operating Systems Design and Implementation.Berkeley,USA:USENIX Association,2016:265-283.
[13] 钟川,陈军.基于精确欧氏局部敏感哈希的改进协同过滤推荐算法[J].计算机工程,2017,34(2):74-78.
[14] WEISS Y,TORRALBA A,FERGUS R.Spectral Hashing[C]//Proceedings of the 21st International Conference on Neural Information Processing Systems.[S.l.]:Curran Associates Inc.,2008:1753-1760.
[15] HE Kaiming,WEN Fang,SUN Jian.K-means hashing:an affinity-preserving quantization method for learning binary compact codes[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Computer Society,2013:2938-2945.
[16] GONG Yunchao,LAZEBNIK S,GORDO A,et al.Iterative quantization:a procrustean approach to learning binary codes for large-scale image retrieval[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(12):2916-2929.