MM-SPARSE 2016

MM-SPARSE 2016

1st Workshop on SPARSITY AND COMPRESSIVE SENSING IN MULTIMEDIA

An IEEE ICME 2016 workshop - July 15, 2016 - Seattle, USA

Download the Call for Papers

News

    • The workshop program is online. Click here for details
    • The deadline for paper submission has been postponed to Mar. 25, 2016
    • Workshop website is online

Important dates

    • Workshop paper submission: Mar. 18, 2016 Mar. 25, 2016
  • Notification of acceptance: Apr. 22, 2016
  • Camera-ready workshop paper submission: May 13, 2016

Technical Program

The workshop will entirely take place in Room Cascade I-B - 8:45am - 5:00pm

8:45am - 8:55am

Welcome and opening remarks

8:55am - 9:55am

Plenary Talk

First-Photon Imaging and Other Imaging with Few Photons - Vivek Goyal (Boston University, USA)

9:55am - 12:00pm (coffee break 10:15am - 10:40am)

Oral Session MM-SPARSE 1. Session chair: Enrico Magli (Politecnico di Torino, Italy)

  • 9:55am: Depth Superresolution Using Motion Adaptive Regularization - Ulugbek Kamilov (Mitsubishi Electric Research Laboratories), Petros Boufounos (Mitsubishi Electric Research Laboratories)
  • 10:15am: Coffee Break
  • 10:40am: BM3D-prGAMP: COMPRESSIVE PHASE RETRIEVAL BASED ON BM3D DENOISING - Chris Metzler (Rice University), Arian Maleki (Columbia University), Richard Baraniuk (Rice University)
  • 11:00am: A Sparse Representation Based Post-processing Method for Improving Image Super-Resolution - Jun Yang (Sun Yat-sen University), Jun Guo (Sun Yat-sen University), Hongyang Chao (Sun Yat-sen University)
  • 11:20am: RAIN REMOVAL VIA SHRINKAGE OF SPARSE CODES AND LEARNED RAIN DICTIONARY - Chang-Hwan Son (Ryerson University), Xiao-Ping Zhang (Ryerson university)
  • 11:40am: SHALLOW SPARSE AUTOENCODERS VERSUS SPARSE CODING ALGORITHMS FOR IMAGE COMPRESSION - Thierry Dumas (INRIA), Aline Roumy (INRIA), Christine Guillemot (INRIA)

12:00pm - 2:00pm

Lunch Break

2:00pm - 3:00pm

Plenary Talk

When less gets more; Using sparsity to resolve computationally intractable audio problems - Paris Smaragdis (University of Illinois at Urbana-Champaign, USA)

3:00pm - 5:00pm (coffee break 3:20pm - 3:40pm)

Oral Session MM-SPARSE 2. Session chair: Petros Boufounos (MERL, USA)

  • 3:00pm: TOOTHPIC: WHO TOOK THIS PICTURE? - Diego Valsesia (Politecnico di Torino), Giulio Coluccia (Politecnico di Torino ), Tiziano Bianchi (Politecnico di Torino), Enrico Magli (Politecnico di Torino)
  • 3:20pm: Coffee Break
  • 3:40pm: SparseHash: embedding Jaccard coefficient between supports of signals - Diego Valsesia (Politecnico di Torino), Sophie Fosson (Politecnico di Torino), Chiara Ravazzi (Politecnico di Torino), Tiziano Bianchi (Politecnico di Torino), Enrico Magli (Politecnico di Torino)
  • 4:00pm: ATOMIC NORM MINIMIZATION FOR MODAL ANALYSIS - Shuang Li (Colorado School of Mines), Dehui Yang (Colorado School of Mines), Michael Wakin (Colorado School of Mines)
  • 4:20pm: ADAPTIVE SALIENCY-BASED COMPRESSIVE SENSING IMAGE RECONSTRUCTION - Ali Akbari (Institut Suprieur d’Electronique de Paris (ISEP) ), Diana Mandache (Institut Suprieur d’Electronique de Paris (ISEP) ), Maria Trocan (Institut Suprieur d’Electronique de Paris (ISEP) ), Bertrand Granado (Pierre et Marie Curie University)
  • 4:40pm: Sparsity and Parallel Acquisition: Optimal Uniform and Nonuniform Recovery Guarantees - Il Yong Chun (Purdue University), Chen Li (University of Science and Technology of China), Ben Adcock (Simon Fraser University)

Confirmed keynote speakers

Paris Smargadis

(University of Illinois at Urbana-Champaign, USA)

(Boston University, USA)


Keynote title: "When less gets more; Using sparsity to resolve computationally intractable audio problems."

Abstract: Audio, like all other signal modalities, has seen its fair share of straightforward sparsity approaches. In this talk I will focus on some different ways that sparsity has been taken advantage of in audio processing problems. Using the context of source separation I will present some probabilistic formulations in which, surprisingly, in order to achieve sparsity one can simply maximize an l2-norm, and then I’ll extend this idea and show how we can scale these methods to very big data sets using sparse manifold methods. In both cases, the use of sparsity helps us reduce problems that are otherwise computationally intractable to relatively easier ones. I’ll then shift gears and describe how imposing sparsity in audio model parameters allows us once more to significantly reduce computational complexity. I will demonstrate that in the context of audio denoising neural network models, and I will show how sparsifying their parameters allows us to reduce the necessary processing to simple bit-wise computations which are well-suited for low-power hardware implementations while retaining the expressibility of a general neural network.

Bio: Paris Smaragdis is faculty at the Computer Science and Electrical & Computer Engineering departments at the University of Illinois at Urbana-Champaign. He is also a senior research scientist at Adobe Research. Prior to that he was a research scientist at MERL and he received his PhD from MIT. His work focuses on computational audition, and machine learning approaches to various audio signal processing problems. He is primarily interested in problems involving mixtures of sounds. He has been selected by the MIT Tech Review as one of the top 35 innovators under 35 years old in 2006, he is an IEEE Fellow, and an IEEE SPS Distinguished Lecturer.

Keynote title: "First-Photon Imaging and Other Imaging with Few Photons"

Abstract: LIDAR systems use single-photon detectors to enable long-range reflectivity and depth imaging. By exploiting an inhomogeneous Poisson process observation model and the typical structure of natural scenes, first-photon imaging demonstrates the possibility of accurate LIDAR with only 1 detected photon per pixel, where half of the detections are due to (uninformative) ambient light and dark counts. I will explain the simple ideas behind first-photon imaging, which include exploitation of transform-domain sparsity. Then I will present related subsequent work that avoids the potential loss of transverse resolution from transform-domain regularization and instead exploits only longitudinal sparsity. This later work yields methods robust to unknown, spatially varying ambient light and able to estimate multiple depths per pixel from about 10 detected photons per pixel.

Bio: Vivek Goyal received the M.S. and Ph.D. degrees in electrical engineering from the University of California, Berkeley, where he received the Eliahu Jury Award for outstanding achievement in systems, communications, control, or signal processing. He was a Member of Technical Staff in the Mathematics of Communications Research Department of Bell Laboratories, a Senior Research Engineer for Digital Fountain, and the Esther and Harold E. Edgerton Associate Professor of Electrical Engineering at MIT. He was an adviser to 3dim Tech, Inc., and is now an Associate Professor of Electrical and Computer Engineering at Boston University.

Dr. Goyal is a Fellow of the IEEE. He was awarded the 2002 IEEE Signal Processing Society Magazine Award, an NSF CAREER Award, and the Best Paper Award at the 2014 IEEE International Conference on Image Processing. Work he supervised won student best paper awards at the IEEE Data Compression Conference in 2006 and 2011 and the IEEE Sensor Array and Multichannel Signal Processing Workshop in 2012 as well as five MIT thesis awards. He currently serves on the Editorial Board of Foundations and Trends and Signal Processing, the Scientific Advisory Board of the Banff International Research Station for Mathematical Innovation and Discovery, the IEEE SPS Computational Imaging SIG, and the IEEE SPS Industry DSP TC. He was a Technical Program Committee Co-chair of Sampling Theory and Applications 2015 and is a permanent Conference Co-chair of the SPIE Wavelets and Sparsity conference series. He is a co-author of Foundations of Signal Processing (Cambridge University Press, 2014).

Organizing committee

Organizing co-chairs

Technical Program Committee

Author information and submission instructions

The following informations are taken from the ICME 2016 website. Please refer to www.icme2016.org for latest updates.

All Workshop papers will have the following commonalities:

  • Length: Papers must be no longer than 6 pages, including all text, figures, and references.
  • Format: Workshop papers have the same format as regular papers. However, they need not be double blind.
  • Review: Reviews will be handled directly by the Workshop organizers.
  • Submission site: https://cmt.research.microsoft.com/ICMEW2016. Papers must be submitted under the track for the appropriate Workshop. Submissions may be accompanied by up to 30 MB of supplemental material following the same guidelines as regular and special session papers.
  • Presentation guarantee: As with accepted Regular and Special Session papers, accepted Workshop papers must be registered by the author deadline and presented at the conference; otherwise they will not be included in IEEE Xplore. A workshop paper may be covered by a workshop-only registration or by a full-conference registration.

Scope

Technology improvements in multimedia content generation have led to a continuous cycle of creation, processing, communication and consumption of information as part of our social interactions and entertainment. This has spurred the development of efficient ways to acquire, process, communicate and store vast amounts of data. One of the most prominent recent breakthroughs is the development of compressed sensing and sparse representations. To-date these techniques have had significant success in numerous areas of signal acquisition, parameter estimation, communication, and machine learning. These developments have found a home in a number of applications, including media retrieval, radar and medical imaging systems, and video compression, just to mention a few.

This workshop’s objective is to bring together leading researchers around the world to present their newest accomplishments, exchange latest experiences, and explore future directions in the important and evolving fields of CS and sparse representations. In particular, it is focused on the dissemination of recent results in multimedia methods and applications that take advantage of these techniques.

We are inviting original submission to this workshop in the form of full-length 6-pages papers, or short 2-pages position papers (these latter are not going to be published in the proceedings). Topics of interest include, but are not restricted to:

  • Theoretical aspects of compressive sensing and sparsity
  • Communication and networking applications
  • CS and sparsity for 5G communication systems
  • CS and sparsity for big data
  • CS and sparsity for media retrieval
  • CS and sparsity for speech, audio, image and video compression
  • CS and sparsity for detection, estimation and machine learning
  • Compressive imaging
  • Distributed applications of compressive sensing
  • Emerging applications of CS and sparsity