20160511-14 顾舒航:Synthesis and analysis sparse representation models for...

报告嘉宾2：顾舒航 （香港理工大学）

报告时间：2016年5月11日（星期三）晚21:00（北京时间）

报告题目：Synthesis and analysis sparse representation models for image modeling

主持人：  刘日升（大连理工大学）

报告摘要：Synthesis and analysis representation are two categories of signal representation models. Based on these two kinds of models, different methods have been proposed to model the prior of natural images, including patch-based and filter-based implementations. In this talk, I first review some representative natural image prior modeling methods and briefly discuss the advantages and disadvantages of them. Then, I introduce an analysis-based and a synthesis based image restoration models, e.g. the weighted nuclear norm minimization (WNNM) model and the convolutional sparse coding super-resolution (CSC-SR) model. WNNM is a 2D analysis model which is able to embed non-local similarity prior into analysis representation models to improve visual quality of restoration results, it achieved stage-of-the-art denoising results both in terms of PSNR and visual quality. CSC-SR utilize convolutional sparse coding to avoid patch aggregation in  previous patch-based synthesis models, it also achieved perception pleasant SR results with stage-of-the-art PSNR index.

参考文献：

[1] S. Gu, L. Zhang, W. Zuo, and X. Feng, “Weighted Nuclear Norm Minimization with Application to Image Denoising,” In 2014 IEEE Conf. on Computer Vision and Pattern Recognition (CVPR 2014).

[2] S. Gu, W. Zuo, Q. Xie, D. Meng, X. Feng, L. Zhang. "Convolutional Sparse Coding for Image Super-resolution," In ICCV 2015.

[3] Weighted Nuclear Norm Minimization and Its Applications to Low Level Vision.  Shuhang Gu, Qi Xie, Deyu Meng, Wangmeng Zuo, Xiangchu Feng, Lei Zhang. Submitted to International Journal of Computer Vision (IJCV)

报告人简介：

Shuhang Gu received the B.E. degree from the School of Astronautics, Beijing University of Aeronautics and Astronautics, China, in 2010, and the M.E. degree from the Institute of Pattern Recognition and Artificial Intelligence, Huazhong University of Science and Technology, China, in 2013. He is currently pursuing the Ph.D. degree with the Department of Computing, The Hong Kong Polytechnic University. His current research interest is  learning and optimization for low level vision. He already published several papers in top conferences including CVPR, ICCV and NIPS.