Call for Papers
Computer Vision and Image Understanding Journal (CVIU)
Special Issue on Spatial Coherence in Visual Motion Analysis
Guest Editors
W. James MacLean, University of Toronto, Canada
Nikos Paragios, Ecole Nationale des Ponts et Chausses, France
David Fleet, University of Toronto, Canada
Motion analysis is a central problem in computer vision, and the past
two decades have seen important advances in this field. However, visual
motion is still often considered on a pixel-by-pixel basis, even though
this ignores the fact that image regions corresponding to a single
object usually undergo motion that is highly correlated. This
independence is often an explicit assumption that is made when
developing computational models. Models in which such independence is
not assumed, for example Markov Random Fields, are typically
computationally expensive, and therefore many times are not the model of
choice. It must be noted that an implicit assumption of spatial
coherence exists in motion models employing region based estimates of
quantities such as image gradients. Further, it is often of interest to
accurately measure the boundaries of moving regions. In the case of
articulated motion, especially human motion, discovering motion
boundaries is non-trivial but an important task nonetheless. Early
approaches focused on measuring motion of either the boundaries or the
interior, but seldom both in unison. In the case of identifying and
tracking independent object motion, such a united approach may be
essential, given the possibly small region subtended by the tracked
object(s). Another related problem is identifying and grouping multiple
disconnected regions moving with similar motions, such as a flock of
geese. In the past several years attempts have been made to include
spatial coherence terms into algorithms for 2- and 3-D motion recovery,
as well as motion boundary estimation.
This special issue will examine the state-of-the-art in techniques for
integrating spatial coherence constraints during motion analysis on
image sequences. While a broad range of topics will be considered,
papers submitted must make a significant contribution to furthering
ability to take advantage of spatial coherence to produce more accurate
and reliable motion estimates. Topics for submitted papers include (but
are not limited to):
o Bayesian models of spatial coherence
o Belief Propagation
o Generative Models
o Markov random field techniques
o Optic Flow
o Recovery of motion boundaries
o Active contours & boundary tracking
o Motion boundary interpretation and occlusion/disocclusion
modeling
o Articulated Motion
o Independent Object Motion
o Visual Tracking
o Layered motion models
o Region segmentation & Motion-based grouping
o Perceptual grouping of pixel motions
o Spatial coherence models for transparency
o Spatial coherence in biological vision
o Human motion analysis
o Use of contextual information in applying spatial coherence
o Local-Parallel computation models for motion
o Graph-Based Methods for Motion Segmentation
All submitted papers will be reviewed according to the guidelines and
standards of the Computer Vision and Image Understanding Journal. We
prefer that the authors submit electronic versions of their papers in
postscript or pdf to W. James MacLean (maclean+cviu@eecg.toronto.edu).
If electronic submission is not possible then five paper copies may be
sent to:
Prof. W. James MacLean,
Edward S. Rogers Sr. Dept of Electrical & Computer Engineering,
University of Toronto
10 King's College Road, Toronto, ON M5S 3G4 Canada
Deadlines
Manuscript submission February 28, 2005
Reviews sent to authors June 30, 2005
Submission of revised manuscripts August 31, 2005
Final accept/reject notification September 30, 2005
Publication date: Fourth quarter 2005
For further information please contact W. James MacLean
( mailto:maclean+cviu@eecg.toronto.edu ).