A SYSTEM FOR IMAGE/VIDEO OBJECT QUERY BY SHAPE - Video Object Extraction and Representation: Theory and Applications - page 132

Figure 6.3. Implicit structure: a) an outline of a dog has implicit structure from
its skeleton, b) an outline cloud does not follow our assumption of implicit support
structure, aa application of our extraction algorithm would result in noise c) a repre-
sentation of a dog wagging its tail, a “happy” dog.

linked to physical properties of the objects. The extraction process of our
shape query system has four components: 1) the assumption of implicit
support structure, 2) the expression of these assumptions via VOS, 3) the
representation of VOS with a graph, and 4) the orders enforced within
the graph. The VOS can be expressed as an ordered tree, but ordered
trees have inherent

robustness problems that motivates our novel data

structure, DAG-Ordered Trees, as a solution in Section 4..

THE ASSUMPTION OF IMPLICIT SUPPORT
STRUCTURE
Through a priori knowledge, we can relate the object shape contour
to physical properties of the object. We can use the physical properties
as the basis of a shape descriptor. As shown in Figure 6.3, we can
induce support structure from a given shape that acts like skeleton for
the object. Our assumptions about the physical object are as follows:

1. Support Structure Existence

If an object protrudes into space, we assume that a time-invariant
internal support structure exists that allows this protrusion (see Fig-
ure 6.3a).