Faculty of Informatics Vienna University of Technology Institute of Computer Aided Automation PRIP Home PRIP Home
Personal tools
You are here: Home Research Current Projects Cognitive Vision Representations of Cognitive Processes

Representations of Cognitive Processes

Understanding and representing processes involved in visual problem solving is crucial when one wants to design adaptable cognitive systems. An adaptable system is required to be able to change the way it behaves (or more precisely it solves problems) when confronted to new problems or when it realizes that a current method is unable to solve a problem.

Humans obviously have this ability. Since the human visual system is the more accessible working cognitive system known up to know, it is of primary importance to study it.

Thus, this component of our research aims at understanding mental mechanisms involved in thinking and solving problems by human beings. This research has two main components: (i) psychophysical experiments, which lead to hypotheses about the underlying mental algorithms; and (ii) computational modeling, which implements the algorithms and tests their efficiency and plausibility as models of the human mind.

Shape Understanding

Many schools in psychophysics claim that humans understand shapes thanks to some simplicity criterion.

In the human visual system, the role of constraints (or model) is preponderant. For example, by closing one eye, a person can still see and reason in three dimensions. Although the human visual system itself is mainly two dimensional, humans immediately interpret images in three dimensions.

Experiments of reconstruction were led showing that, using settings similar to the human visual system, reconstruction of objects with one image and constraints is much better than using only a stereoscopic image.

Local assumptions (like convexity of corner) and a priori knowledge seem to lead together to a correct interpretation of shape at a global level. Thus, it seems that recognition of shape is done at a global level thanks to local processes involving a priori constraints.

Thinking & Solving Problems

Perception, thinking, and decision making, are cognitive functions that allow a human being to obtain and analyze information about the environment. It is commonly acknowledged that the human mind accomplishes its cognitive goals very efficiently (perhaps even optimally). However, from a computational perspective, these cognitive tasks are extremely difficult: at the present, no artificial system comes even close to the level of performance demonstrated by the human mind.

Studying well known optimization problems like the euclidean traveling salesman problem has in fact a larger scope, as most of the problems, in AI and computer vision, can be stated as combinatorial optimization problems.

The traveling salesman problem (tsp) is known to belong to the class of NP complete problems. Zygmunt Pizlo's recent work shows that humans are able to find optimal, or close to optimal tours very quickly. Based on results of psychophysical experiments, he proposed a computational model which takes the form of a pyramid algorithm. The current research concentrates on analyzing computational aspects of tsp, and on elaborating the existing algorithm.

A program used in order to experiment tsp problem solving can be downloaded here.

Results of this cooperation are presented in [PSS+06].

External Collaborations

  1. Dr. Zygmunt Pizlo, Visual Perception Laboratory, Purdue University, USA

Related Bibliography

  1. Y.G. Leclerc, Constructing Simple Stable Descriptions for Image Partitioning; International Journal of Computer Vision, 1989.
  2. Pizlo, Z., Perception viewed as an inverse problem. Vision Research, 41, page 3145-3161, 2001.
  3. S.M.Graham, A.Joshi & Z. Pizlo (2000) The Traveling Salesman Problem: a hierarchical model. Memory & Cognition 28, 1191-1204.
  4. See the List of PRIP publications created during this project.