DanskDTU.dkIndexContactPhonebookPortalenDTU Alumni
DTU

 

 

Summer School on

Advances in Image, Motion and Video Coding and Processing

 

In Copenhagen, Denmark, August 22-26, 2011

The Summer School is held at DTU, Lyngby Campus, Building 101.

 

Motion and optical flow are basic elements of video analysis and video coding. Video coding is a prominent application of motion analysis and image and video coding provides a good setting for benchmarking and improving image and video models and techniques.

 

Standard methods for image and video compression use information theoretic ap- proaches. E.g. transforming the data representation to one where correlation and redundancy appear more clearly, and use this new representation with lossy or lossless encoding schemes. For video there is a considerable temporal redundancy of successive frames that can be exploited. In state of the art video codecs (e.g. H.264/AVC) this redundancy is typically utilized at the encoder side using block-based motion-compensated methods to compress the video data. An alternative and newer type of video coding is distributed video coding, which builds on information theoretic results showing that the standard en-coding/decoding setup can be reversed in the sense that the computationally hard work of using e.g. temporal redundancy, can be done at decoder side rather than at the encoder. Distributed video coding is one instance of distributed source coding.

 

In image analysis content of interest is usually characterized directly and often in a continuous setting (as opposed to video coding). Content of interest can be edges, T-junctions, scale-space structures, etc. Recent research has shown that a proper sparse description of these features can lead to efficient compression algorithms, with reconstructions governed via carefully designed partial differential equations, via variational formulations.

 

The same is true for video analysis, where problems, when embedded in the continuous setting, will often yield much simpler forms as variational formulations and resulting partial differential equations or problems in convex analysis, while they provide some of the most accurate motion recovery techniques. They could thus be beneficial to the Video Coding community.

 

The areas of image and video coding and analysis have a large number of similar goals, but often very different approaches to reaching them. In this summer school, we will let the two communities meet for exchange of ideas on related topics.

 

The PhD course will consist of lectures given by members of the research groups in Coding at DTU Fotonik, Technical University of Denmark and in Image Analysis at DIKU, Copenhagen University as well as invited speakers.

 

Keynote lecturers

Prof. Xiaolin Wu, McMaster University, Canada

Prof. Joachim Weickert Saarland University, Germany

Prof. Fernando Pereira, Instituto Superior Tecnico, Portugal

Prof. Horst Bischof, TU Graz, Austria 

 

 

 

Keynote lectures:

 

Xiaolin Wu  High-speed video acquisition by multi-camera coded exposure
   Optimizing backlight dimming displays
   Soft decoding of near-lossless coded images

Joachim Weickert  

 PDE based image coding I

   PDE based image coding II
Fernando Pereira  Distributed Video Coding Abstract
   Trends on visual coding (including 3D video coding)
Horst Bishof  Variational Methods
   On-line learning

 

 

Topics

  • Video Coding:
    • Distributed video coding
    • 3D video coding
    • New steps in video coding standardization
    • High-speed video acquisition by multi-camera coded exposure  
  • Variational techniques: 
    • Variational Optical Flow on the GPU
    • Interactive Variational Segmentation
    • Variational Superresolution
    • Variational Stereo
    • On-line learning based tracking methods  
  • Image compression:
    • PDE-Based Image Compression
    • Soft decoding of near-lossless coded images  
  • Optimal local backlight dimming algorithms for display
 

 

Program at a glance

 

Full Summer School program

 

 

Credits

PhD ECTS: You will get 2.5 ECTS for participating. You can get additional 2.5 ECTS if you present a poster at the summer school and write a small report on how the topics covered in the course relate to your own PhD project. That is in total you can get 5 ECTS points.

 

Organization

The summer school is organized in collaboration between the Image Group, Department of Computer Science, University of Copenhagen and the Coding and Visual Communication Group, DTU Fotonik, Technical University of Denmark.

 

Organizing committee

Søren Forchhammer (DTU Fotonik),

Lars Lau (DIKU),

 

Registration and practical matters

The registration fee is 2000 dkr. payable upon acceptance to the summer school.

Payment is made here.

 

For registration send an email to Britt Boding (DTU Fotonik),

The registration covers lunch, refreshments, coffee, and dinner during course of the summer school.

 

Deadline for registration has passed.

 

 

Lecturers

Xiaolin Wu

Joachim Weickert

Fernando Pereira

Horst Bischof

Mads Nielsen

Søren Forchhammer

Francois Lauze

Jari Korhonen  

 

 

 

Register by sending an

e-mail to

 

Payment is made here.

 

_________________

Keynote lecturers

 

Prof. Xiaolin Wu,

McMaster University, Canada

 

Prof. Joachim Weickert,

Saarland University, Germany

 

Prof. Fernando Pereira,

Instituto Superior Tecnico, Portugal

 

Prof. Horst Bischof,

TU Graz, Austria