ROBUST REPRESENTATION OF SHAPE WITH DAGS - Video Object Extraction and Representation: Theory and Applications - page 126

7. CONCLUSION
This chapter applies Directed Acyclic Graphs (DAGs) to a large class
of pattern recognition problems and other recognition problems where
the data has a linear ordering. The datastreams are DAG-coded for
robust partitioning. The similarity of two datastreams can be mani-
fested as the path matching score of the two corresponding DAG o s.
This chapter also presents the DAG-Compare algorithm an efficient and
robust dynamic programming algorithm for comparisons of two DAG o s.
Since the DAG-Coding provides a robust partitioning process, it can be
applied recursively to create a novel system architecture. Not discussed
in this chapter, the DAG structure also allows adaptive restructuring,
leading to a novel approach to neural information processing [Lin and
Kung, 1998b]. DAG-Coding may also be applied to any datastream
where a complex partitioning aids the recognition process. Although
the majority of the initial work was done on a test bed of handwriting
recognition problems, we are of opinion that DAGs can be used on any
datastream where a complex partitioning process can aid in recognition.
In Chapter 6, we have expanded the DAG o data structures with recur-
sion to implement shape query system in support of MPEG-7 standard.