Summary: Automatic video segmentation using
spatiotemporal T-junctions
Nicholas Apostoloff Andrew Fitzgibbon
University of Oxford Microsoft Research Ltd.
http://www.robots.ox.ac.uk/vgg http://research.microsoft.com/mlp
Abstract
The problem of figure­ground segmentation is of great importance in both
video editing and visual perception tasks. Classical video segmentation algo-
rithms approach the problem from one of two perspectives. At one extreme,
global approaches constrain the camera motion to simplify the image struc-
ture. At the other extreme, local approaches estimate motion in small image
regions over a small number of frames and tend to produce noisy signals that
are difficult to segment. With recent advances in image segmentation show-
ing that sparse information is often sufficient for figure­ground segmentation
it seems surprising then that with the extra temporal information of video, an
unconstrained automatic figure­ground segmentation algorithm still eludes
the research community. In this paper we present an automatic video seg-
mentation algorithm that is intermediate between these two extremes and
uses spatiotemporal features to regularize the segmentation. Detecting spa-
tiotemporal T-junctions that indicate occlusion edges, we learn an occlusion

Source: Apostoloff, Nicholas - Department of Engineering Science, University of Oxford
Zisserman, Andrew - Department of Engineering Science, University of Oxford

Collections: Computer Technologies and Information Sciences; Engineering