好文作者面授招–20160113

【16-02期VALSE Webinar活动】

Xionghongkai报告嘉宾1:熊红凯(上海交通大学)
报告时间:
2016年1月13日(星期三)晚20:00(北京时间)
报告题目: 
Structured Modeling and Learning in Generalized Data Compression and Processing [Slides]
主持人: 
卢孝强(中科院西光所)
报告摘要:
To represent increasingly massive volume of heterogeneous data with complex structures, classical context modeling maps the context space into a parameter set constructed from statistics of the source. It is parametrically represented in a probabilistic framework that concerns the structure of context models with probability assignment. The typical algorithms are rooted in the finite context statistics incorporating variable-order Markov models, where the optimal context model is selected by adaptively estimating its order. From multi-directional extension and combinatorial structuring of contexts, this talk address learning structured probabilistic model which is composed of utilizing probabilistic graphical model to represent the complex structures in heterogeneous data, making model-based inference for learning, and adopting reasoning algorithms to optimize learning process.

By learning and optimizing, structured prediction model utilizes the complex structure to make prediction for sets of prediction tasks simultaneously. Such method generates less information than the combination of individual predictions, which is suitable for heterogeneous data compression. Namely, generalized context modeling (GCM) is established to capture complex structures in heterogeneous data. It extends the suffix of predicted subsequences in classic context modeling to arbitrary combinations of symbols in multiple directions. To address the selection of contexts, GCM constructs a model graph with a combinatorial structuring of finite order combination of predicted symbols as its nodes. The estimated probability for prediction is obtained by weighting over a class of context models that contain all the occurrences of nodes in the model graph. Moreover, separable context modeling in each direction is adopted for efficient prediction. To find optimal class of context models for prediction, the normalized maximum likelihood (NML) function is developed to estimate their structures and parameters, especially for heterogeneous data with large sizes. Furthermore, it is refined by context pruning to exclude the redundant models. Such model selection is optimal in the sense of minimum description length (MDL) principle, whose divergence is proven to be consistent with the actual distribution. It is shown that upper bounds of model redundancy for GCM are irrelevant to the size of data. GCM is validated in an extensive field of applications, e.g., Calgary corpus, executable files, and genomic data, lossless image coding and intra-frame video coding.
参考文献:
[1] W. Dai, H. Xiong*, J. Wang, S. Cheng, Y. F. Zheng, “Generalized Context Modeling with Multi-Directional Structuring and MDL-based Model Selection for Heterogeneous Data Compression”, IEEE Transactions on Signal Processing, vol. 63, no. 21, pp. 5650-5664, Nov. 2015.
[2] W. Dai, H. Xiong*, J. Wang, Y. F. Zheng. “Structured Set Prediction Modeling with Max-Margin Markov Network for Lossless Image Coding,” IEEE Transactions on Image Processing, vol. 23, no. 2, pp. 541-554, Feb. 2014.
[3] W. Dai, H. Xiong*, X. Jiang, C. W. Chen, “Structured Set Intra Prediction with Discriminative Learning in Max-Margin Markov Network for High Efficiency Video Coding”, IEEE Transactions on Circuits and Systems for Video Technology, vol. 23, no. 11, pp. 1941-1956, Nov. 2013.

报告人简介:
Hongkai Xiong is currently a distinguished Professor in the Department of Electronic Engineering, Shanghai Jiao Tong University (SJTU). Since he received his Ph.D. degree from SJTU in 2003, he has been with Department of Electronic Engineering in SJTU. During 2007-2008, he was a research scholar in the Department of Electrical and Computer Engineering of Carnegie Mellon University (CMU). From 2011 to 2012, he was a scientist with the Department of Biomedical Informatics at the University of California, San Diego (UCSD).

Dr. Xiong’s research interests include signal processing, multimedia communication, source coding and computer vision. He published over 140 refereed journal and conference papers. His research projects are funded by NSF, QUALCOMM, MICROSOFT, and INTEL. He was the recipient of the Best Student Paper Award at the 2014 IEEE Visual Communication and Image Processing (IEEE VCIP’14), the Best Paper Award at the 2013 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (IEEE BMSB’13), and the Top 10% Paper Award at the 2011 IEEE International Workshop on Multimedia Signal Processing (IEEE MMSP’11). He served as TPC members for prestigious conferences such as ACM Multimedia, ICIP, ICME, and ISCAS.

In 2014, Dr. Xiong was granted National Science Fund for Distinguished Young Scholar and Shanghai Youth Science and Technology Talent as well. In 2013, he was awarded a recipient of Shanghai Shu Guang Scholar. From 2012, he is a member of Innovative Research Groups of the National Natural Science. In 2011, he obtained the First Prize of the Shanghai Technological Innovation Award for “Network-oriented Video Processing and Dissemination: Theory and Technology”. In 2010 and 2013, he obtained the SMC-A Excellent Young Faculty Award of Shanghai Jiao Tong University. In 2009, he was awarded a recipient of New Century Excellent Talents in University, Ministry of Education of China. He is a senior member of the IEEE (2010).

熊红凯,上海交通大学特聘教授、博士生导师,国家杰出青年科学基金获得者,教育部新世纪优秀人才,上海市曙光学者,上海市青年科技英才,国家自然科学基金委创新研究群体成员,上海市技术发明奖一等奖(排名第1)。IEEE高级会员、ACM会员。
2003年,在上海交通大学获得通信与信息系统专业博士,留校工作。2007~2008年,在美国卡内基梅隆大学CMU电气与计算机工程系ECE,担任研究员;2011~2012,在美国加州大学圣地亚哥分校(UCSD)生物医学信息中心,担任Scientist。主要研究方向为:信息论与编码、信号处理、多媒体通信、生物医学信息。在相关国际著名期刊和学术会议上,共发表SCI期刊论文40余篇,其中IEEE Trans汇刊30多篇,EI收录论文140多篇。共主持:国家自然科学基金重点项目2项、面上项目2项、海外及港澳学者合作项目3项、863项目等。
2011年,获“上海市技术发明奖”一等奖(第一完成人,“网络化的视频媒体处理与适配分发关键技术与核心系统”)。2014年,指导学生获IEEE VCIP最佳学生论文奖;2013年,获IEEE BMSB(多媒体通信与广播)最佳论文奖;2011年,获IEEE MMSP(多媒体信号处理)Top 10 %论文奖。2010与2013年,2次入选上海交通大学“SMC-A类晨星青年学者计划”。在上海交通大学组建“图像-视频-多媒体”IVM实验室(http://ivm.sjtu.edu.cn),是IEEE多媒体通信技术委员会委员、IEEE视觉信号处理与通信委员会委员。

Fengjianjiang报告嘉宾2:冯建江(清华大学)
报告时间: 2016年1月13日(星期三)晚21:00(北京时间)
报告题目:Statistical Model Based Detection, Segmentation, and Enhancement with Applications in Fingerprint Recognition [Slides]
主持人: 韩琥(中科院计算所)
报告摘要:
Latent fingerprints have been used by law enforcement agencies to identify suspects for a century. However, because of poor image quality and complex background noise, latent fingerprints are routinely identified relying on features manually marked by human experts in practice. A large number of latent fingerprints cannot be treated in time due to lacking well trained experts, highlighting the need for fully automatic systems. We propose a systematic algorithm for latent fingerprint detection, segmentation, and orientation field estimation, without any manual markup. The proposed algorithm is based on a statistical model of fingerprint orientation fields, localized orientation dictionaries. Multiple potential latent fingerprints are detected using a sequential pose estimation algorithm. Then, the full orientation field and confidence map of each detected fingerprint are estimated based on localized dictionaries lookup. Finally, the boundary of each latent fingerprint is delineated by analyzing its confidence map. Experiments on a multi-latent fingerprint database and the challenging NIST SD27 latent fingerprint database show the effectiveness of the proposed algorithm.
参考文献:
[1] Xuanbin Si, Jianjiang Feng, Jie Zhou, Yuxuan Luo: Detection and Rectification of Distorted Fingerprints. IEEE Trans. Pattern Anal. Mach. Intell. 37(3): 555-568 (2015).
[2] Xiao Yang, Jianjiang Feng, Jie Zhou: Localized Dictionaries Based Orientation Field Estimation for Latent Fingerprints. IEEE Trans. Pattern Anal. Mach. Intell. 36(5): 955-969 (2014).
[3] Jianjiang Feng, Jie Zhou, Anil K. Jain: Orientation Field Estimation for Latent Fingerprint Enhancement. IEEE Trans. Pattern Anal. Mach. Intell. 35(4): 925-940 (2013).
[4] Jianjiang Feng, Anil K. Jain: Fingerprint Reconstruction: From Minutiae to Phase. IEEE Trans. Pattern Anal. Mach. Intell. 33(2): 209-223 (2011).
[5] Xuanbin Si, Jianjiang Feng, Jie Zhou: Enhancing latent fingerprints on banknotes. IJCB 2014: 1-8.
报告人简介:
Jianjiang Feng is an associate professor in the Department of Automation at Tsinghua University, Beijing. He received the BS and PhD degrees from the School of Telecommunication Engineering, Beijing University of Posts and Telecommunications, China, in 2000 and 2007, respectively. From 2008 to 2009, Dr. Feng was a postdoctoral researcher in the PRIP Lab at Michigan State University. Dr. Feng’s research interests include fingerprint recognition and computer vision, and he has 16 years research experience in fingerprint recognition, and is the PI of two NSFC award. His fingerprint recognition technologies have been licensed to top biometrics companies from all over the world. Dr. Feng has published over than 40 peer-reviewed academic papers on top journals and conferences, including 9 regular papers on IEEE T-PAMI, and applied 20+ patents in China and 4 patents in US. Dr. Feng received three best paper awards from the Int’l Conference on Biometrics (ICB), two second-place awards of natural science from the Chinese Ministry of Education, and the first-place award of Advance of Science and Technology from the Chinese Institute of Electronics. Dr. Feng is an associate editor of Image and Vision Computing, and area chair of ICB.

冯建江,清华大学自动化系副教授,主要研究方向为图像处理与模式识别。分别于2000年和2007年在北京邮电大学获得本科与博士学位,2008-2009年在美国密歇根州立大学Anil K. Jain研究组担任博士后研究员,从事指纹识别研究工作。在指纹识别领域有16年研究经历,承担国家自然科学基金项目2项;在计算机视觉与生物特征识别领域发表学术论文40多篇,其中IEEE T-PAMI长文9篇;先后申请国内发明专利20余项、美国发明专利4项。所研发的整形指纹检测技术、低质量指纹增强技术授权国内外顶尖企业。先后获International Conference on Biometrics(ICB)等会议最佳论文奖3次、教育部自然科学类二等奖2次以及电子学会科技进步类一等奖。目前担任《Image and Vision Computing》编委(2014年至今)、ICB领域主席(2014年至今)。

(Visited 534 times, 1 visits today)