First IEEE Computer Society Workshop on

Perceptual Organization in Computer Vision

June 26 (post CVPR '98) 
Santa Barbara, California




As recognized by the Gestalt school, the importance of perceptual organization (PO) in human vision cannot be overestimated; it imparts both efficiency and robustness to the visual process. Since early demonstrations in the 1980s underscored its usefulness in object recognition, the computer vision community has seen various applications of PO in artificial vision systems such as in stereo matching, model indexing, contour completion, figure-ground segmentation, change detection, and more. Indeed, it can be argued that a reasonable computational model of perception can be built around the notion of repeated detection and classification of organized structure. PO represents much of the often overlooked intermediate level processing in computer vision systems. So, despite these observations, the full potential of PO in artificial vision systems is yet to be realized. We intend to bring together researchers in perceptual organization in an attempt to crystallize the concepts being explored. In addition to the presentation of new ideas, this would provide a forum to debate the role(s) of perceptual organization in artificial vision systems and thus help to outline future research directions. 


The workshop will be divided into two parts. The first will consist of high quality technical papers describing new or ongoing work in perceptual organization. The second part will consist of structured group discussions. The attendees will break up into groups to consider the challenges in developing perceptual organization methods in vision such as: Types of groups and grouping hierarchies; Efficiency issues; The role of grouping in object recognition, low and intermediate level vision, and motion and stereo; Applications; and Performance evaluation, measures, and standards. The organizing committee will set and disseminate the precise list of questions well before the workshop to provide participants ample time to prepare. We also intend to provide a set of standard images well in advance of the workshop to allow participants to apply their techniques to a common data set for comparative purposes. 


This will be the first workshop focused on perceptual organization in computer vision. We hope to ascertain the present state of the art in perceptual organization and to shape new research directions. The accepted papers will be disseminated among the participants as an informal proceedings. A summary of the panel discussion will be available online.  The output of this workshop is a special issue on Perceptual Organization in Computer in the Computer Vision and Image Understanding Journal (Oct 1999 issue). 

Sudeep Sarkar

Last modified: Mon Jun 1 12:58:49 EDT 1998