International Workshop on The Representation and Use of Prior Knowledge in Vision In conjunction with ECCV 2006, Graz, Austria

Workshop venue, scientific scope, workshop topics, and planned publication

We invite you to participate to the international workshop on the representation and use of prior knowledge in vision (WRUPKV) which will be held on Saturday, May 13, 2006 in Graz, Austria, in conjuction with the ninth European conference on computer vision (ECCV 2006).

It is planned to publish the accepted papers as post-proceedings in Springer LNCS.

For the last decades, computer vision has developed theories for a variety of visual processes. This has been done along the idea that the visual perception problem can be solved within the framework of the information-processing theory: first by analyzing what information should be extracted from images and videos, then building a mathematical model of how this information can be computed, and followed by the design of computational models with their associated implementations.

In general, these approaches proceed bottom-up, in the sense that they attempt to map images and videos onto more abstract descriptions. Interesting enough, while this is consistent with neurophysiolgical findings that the visual field is mapped onto the visual areas in the cortex, it fails to take into account the fact that there is massive feedback between these cortical areas and that visual information travels both forward and backwards.

Therefore and in spite of tremendous successes, the question of which prior knowledge the visual processes -- whether computational or biological -- should use, has remained unanswered. How prior knowledge is represented and how is it used within a top-down flow of information? How top-down and bottom-up processes interact and how do they influence each other?

The workshop's chairs and programme committee members believe that the following topics should be addressed in order to contribute to the advancement of the state of the art in both computer vision and the computational modelling of biological vision:

• 2-D and 3-D shape representation and recognition,
• Figure-ground segmentation,
• Object recognition,
• Image and/or scene recognition,
• Motion segmentation and action/gesture recognition,
• Attentional mechanisms, and
• Visual exploration

These topics could be addressed within one (or several) of the following methodological paradigms:

• Bayesian inference,
• Bayesian decision theory,
• Hidden Markov models,
• Belief propagation,
• Spectral-graph theory and spectral clustering,
• Graph partitioning, and
• Graph matching

Each submitted paper will go through a double-blind review by three programme committee members. Accepted papers will be published both as:

• CD-Rom proceedings, as well as a
• Springer LNCS post-proceedings publication.