基于特征融合和曼哈顿量化的视频哈希学习方法

南京大学学报(自然科学版) ›› 2016, Vol. 52 ›› Issue (4): 705–.

基于特征融合和曼哈顿量化的视频哈希学习方法

聂秀山^1，4，王舒婷¹，尹义龙^2，3*

出版日期:2016-07-24 发布日期:2016-07-24
作者简介: 1.山东财经大学计算机科学与技术学院，济南，250014；2.山东大学计算机科学与技术学院，济南，250100；3．山东省金融信息工程技术研究中心，济南，250100；4.山东省数字媒体重点实验室，济南，250014
基金资助:

基金项目：NSFC－广东联合基金重点支持项目(U1201258)，国家自然科学基金(61573219)，山东省自然科学杰出青年基金(JQ201316)

收稿日期：2016－03－25
*通讯联系人，Email：ylyin@sdu.edu.cn

Video hash learning based on feature fusion and Manhattan quantization

Nie Xiushan^1，4，Wang Shuting¹，Yin Yilong^2，3*

Online:2016-07-24 Published:2016-07-24
About author:1.School of Computer Science and Technology，Shandong University of Finance and Economics，Ji’nan，250014，China；2．School of Computer Science and Technology，Shandong University，Ji’nan，250100，China；
3．Shandong Engineering and Technology Center of Financial Informatization，Ji’nan，250100，China；4．Digital Media Technology Key Laboratory of Shandong Province，Ji’nan，250014，China

摘要/Abstract

摘要： 当前信息时代，随着计算机和多媒体技术的发展，在互联网尤其是移动互联网中，因视频数据结构复杂，特征维度高，其存储、传输和检索都面临着巨大的挑战，视频哈希学习是解决上述挑战的重要方法之一，已成为多媒体处理领域的研究热点．现有方法主要是利用视频不同特征构造视频哈希，但不同特征存在关联关系，为充分利用视频不同特征之间的关联关系，克服传统视频哈希编码的局限性，提出一种基于特征融合和曼哈顿量化的视频哈希学习方法．该方法首先提取视频的全局、局部和时域特征，并利用张量分解理论实现不同特征的融合，获取视频融合特征表示．然后使用曼哈顿量化对视频融合特征进行量化学习编码，得到视频哈希序列．与传统视频哈希算法相比，该方法不仅充分利用了多特征之间的关联互助关系，而且对原始视频特征的不同维度分别进行编码，较好的保持了原始特征之间的结构相似性．实验结果显示，该方法具有较好的性能．

Abstract: With the development of computer and multimedia technologies，video storage，transmission and retrieval are facing a huge challenge in the Internet especially the mobile Internet，due to the complex structure and high dimension of the video.Video hash learning is one of the important ways to solve the challenge，and it becomes one of the hot topics in the field of multimedia processing.As known，the existing methods generate video hashes using different types of features.In fact，there are potential relationships among different types of video features.Therefore，to make full use of the relationships among different video features and overcome the limitations of traditional video hashing methods，we proposed a video hash learning method based on feature fusion and Manhattan quantization in this paper.In the proposed method，the global，local and temporal features are firstly extracted from the video content，and the video clip is considered as a thirdorder tensor.Then，the tensor decomposition，which is popularly applied in multidimensional data processing，is used to fuse the global，local and temporal features.The three loworder tensors are obtained after tensor decomposition，and we concatenate them as the fusion representation of video content.Subsequently，the fused video feature is quantified by Manhattan quantization to get the video hash codes，which are used to construct the final video hash.Compared with the traditional video hashing methods，the proposed method not only makes full use the relationship among different video features，but also achieves the goal of coding with different dimensions respectively，which can well preserve the structural similarity among different video features.Two kinds of experiments are conducted to evaluate the performance of the proposed method，and the results show that the proposed method has a good performance compared with the existing methods.

聂秀山^1，4，王舒婷¹，尹义龙^2，3*. 基于特征融合和曼哈顿量化的视频哈希学习方法[J]. 南京大学学报(自然科学版), 2016, 52(4): 705–.

Nie Xiushan^1，4，Wang Shuting¹，Yin Yilong^2，3*. Video hash learning based on feature fusion and Manhattan quantization[J]. Journal of Nanjing University(Natural Sciences), 2016, 52(4): 705–.

参考文献

[1]　李武军，周志华．大数据哈希学习：现状与趋势．科学通报，2015，60(5－6)：485－490.(Li W J，Zhou Z H．Learning to hash for big data：Current status and future trends.Chinese Science Bulletin，2015，60(5－6)：485－590.)
[2] Kong W H，Li W J，Guo M Y．Manhattan hashing for largescale image retrieva.In：Proceedings of ACM Conference on Research and Development in Information Retrieval(SIGIR)．New York：ACM Press，2012，45－54.
[3] Hampapur A，Hyun K H，Bolle R M．Comparison of sequence matching techniques for video copy detection.In：Proceedings of SPIE，Storage and Retrieval for Media Databases.San Jose：SPIE Press，2002，4676：194－201.
[4] Barrios J M，Bustos B．Competitive contentbased video copy detection using global descriptor.Multimedia Tools and Applications，2013，62(1)：75－110.
[5] Xiang S J，Yang J Q，Huang J W．Perceptual video hashing robust against geometric distortions.Science Chinainformation Sciences，2012，55(7)：1520－1527.
[6] Cosku B，Bulent S，Nasir M．Spatiotemporal transform based video hashing.IEEE Transactions on Multimedia，2006，8(6)：1190－1208.
[7] Esmaeili M M，Fatourechi M，Ward R K．A robust and fast video copy detection system using contentbased fingerprinting.IEEE Transactions on Information Forensics and Security，2011，6(1)：213－226.
[8] Herbert B，Andreas E，Tinne T．SURF：Speededup robust features.Computer Vision and Image Understanding，2008，110(3)：346－359.
[9] Yang G，Chen N，Jiang Q．A robust hashing algorithm based on SURF for video copy detection.Computers & Security，2012，31(1)：33－39.
[10] Lee S，Yoo C D．Robust video fingerprinting for content based video identification.IEEE Transactions on Circuits Systems and Video Technology，2008，18(7)：983－988.
[11] 赵玉鑫，刘光杰，戴跃伟等．基于局部排序的视频复制检测．计算机辅助设计与图形学学报，2009，21(9)：1339－1343.(Zhao Y X，Liu G J，Dai Y W，et al.Video detection based on local ordinal.Journal of Computer Aided Design and Computer Graphics，2009，21(9)：1339－1343.)
[12] Wei Z K，Zhao Y，Zhu C，et al.Frame fusion for video copy detection.IEEE Transactions on Circuits Systems and Video Technology，2011，21(1)：15－28.
[13] Raginsky M，Lazebnik S．Localitysensitive binary codes from shiftinvariant kernels.In：Proceedings of the Annual Conference on Neural Information Processing Systems.Vancouver：Curran Associates Inc.，2009，1509－1517.
[14] Weiss Y，Torralba A，Fergus R．Spectral hashing.In：Proceedings of the Annual Conference on Neural Information Processing Systems.Vancouver：Curran Associates Inc.，2008，1753－1760.
[15] Zhang D，Wang J，Cai D．et al.Selftaught hashing for fast similarity search.In：Proceedings of the ACM Special Interest Group on Information Retrieval.New York：ACM Press，2010，18－25
[16] Li M，Monga V．Robust video hashing via multilinear subspace projections.IEEE Transactions on Image Processing，2012，21(10)：4397－4409.
[17] Morten M．Applications of tensor(multiway array)factorizations and decompositions in data mining.Wiley Interdisciplinary Reviews：Data Mining and Knowledge Discovery，2011，1(1)：24－40.
[18] Nion D，Lathauwer L D．An enhanced line search scheme for complexvalued tensor decompositions.Application in DSCDMA，Signal Process，2008，88(3)：749－755.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed