Apply GPCA to Motion Segmentation - Intelligent Video Event Analysis and Understanding - page 21

Apply GPCA to Motion Segmentation
Hongchuan Yu and Jian J. Zhang *
Abstract. In this paper, we present a motion segmentation approach based on the
subspace segmentation technique, the generalized PCA. By incorporating the cues
from the neighborhood of intensity edges of images, motion segmentation is
solved under an algebra framework. Our main contribution is to propose a post-
processing procedure, which can detect the boundaries of motion layers and fur-
ther determine the layer ordering. Test results on real imagery have confirmed the
validity of our method.
1 Introduction
An important problem in computer vision is to segment moving objects of a scene
from a video source, and partly recover the structure or motion information, such
as foreground and background. With widespread demands on video processing,
motion segmentation has found many direct applications. Video surveillance sys-
tems seek to automatically identify people, objects, or activities of interest in a
variety of environments with a set of stationary cameras. Motion segmentation can
provide low level motion detection and region tracking cues. Another relatively
new application is markerless motion capture for computer animation. It aims to
estimate the human body configuration and pose in the real world from a video by
locating the joint positions over time and extracting the articulated structure.
Motion segmentation is expected to partly recover the structure and motion in-
formation of moving objects from a mutually occluded scene. This includes the
following main tasks, (1) labeling the regions of a motion layer segmentation, i.e.
pixels are assigned to several motion layers; (2) finding their motion, e.g. each layer
has its own smooth flow field while discontinuities occur between layers; (3) deter-
mining the layer ordering, as the different layers might occlude each other. But mo-
tion segmentation is not equivalent to object tracking. Roughly speaking, object