In many problems involving advanced computer-aided production, for example,
navigation of robots in production and maintenance, automatic inspection a
nd testing of manufactured parts, etc., automatic segmentation of video sig
nals into constituent objects is essential. In the current work, we present
a formulation for unsupervised video segmentation and object tracking. Thi
s formulation does not require the supervision of a human user. Each frame
in the video is partitioned into different segments and the segments are co
mbined to form object traces. We provide an algorithm that simultaneously p
artitions a video frame and obtains the parameters of the underlying classe
s. The problem of partitioning each frame is posed as a joint estimation of
the partition and class parameter variables. By incorporating the partitio
n information of the previous frame into the segmentation process of the cu
rrent frame, our method implicitly uses the temporal information. Experimen
tal results show that our method succeeds in capturing the object classes e
ven when the objects undergo translations and rotations not in the plane of
the image. (C) 2000 Elsevier Science B.V. All rights reserved.