Digital Signal Processing Reference
In-Depth Information
the most efficient way for MVI/V segmentation, this still remains a challenging
problem in the research community because of the insufficient accuracy and
robustness.
In this chapter, we focus on the techniques of object-based image segmentation
and video tracking from MVI/V. The rest of the chapter is divided into two parts:
multiview image segmentation and multiview video tracking. In the reminder, image
segmentation from multiview images is discussed in Sect.
5.2
, with compressive re-
view on the current algorithms and description of the proposed method. Section
5.3
addresses the overview on the existing algorithms following our proposed method
on the topic of video segmentation and tracking from multiview video. Finally, con-
clusions are drawn in Sect.
5.4
.
5.2
Multiview Image Segmentation
A series of multiview images (MVIs) can be either simultaneously captured by mul-
tiple cameras, or more commonly and economically, collected by a single camera at
different viewpoints and time instances. According to the visual content, multiview
image segmentation can be grouped in to region-based segmentation and object-
based segmentation. The categorical overview of multiview image segmentation
approaches is shown in Fig.
5.1
. Region-based segmentation aims to cluster the
perceptually similar pixels in the image into homogenous regions, while object-
based segmentation tries to extract the meaningful object and separate foreground
from background. Region-based segmentation focuses on the interpreting and un-
derstanding of the whole scene which is represented by semantically and geomet-
rically consistent partitions as shown in Fig.
5.2
b. On the contrary, object-based
segmentation pays more attention to access and manipulate the OOIs, and the
extracted objects are highlighted in the foreground mask as shown in Fig.
5.2
d.
Fig. 5.1
Overview of multiview image segmentation approaches
