|
ABSTRACT
In this thesis, we try to solve two hand-drawing relating problems They
are hand-drawing symbol recognition and hand-drawing figure similarity
measurement. The solutions to these two problems let computer be more
complaisant and more convenient to the users. In the first task, the
objects to be recognized here are symbols which are composed of
fundamental figures, such as "upward arrow", "delete" etc. A new
scheme for hand-drawing symbol recognition using fundamental figure
extraction algorithm has been suggested. The proposed scheme imposes
no restriction on stroke sequence and scale. In the second task, the
goal is to decide the similarity between two hand-drawing figures. A
data structure called hierarchical interrelated tree (HIT) is proposed
to represent a hand-drawing object. The hierarchical interrelation tree
is effective to decompose the figure and reduce the evaluation time of
similarity measurement. Besides, a novel method is proposed to measure
the similarity between two HITs. It is consistent with the human visual
interpretation and gives a better way for evaluating the similarity
between two hand-drawing figures.
In addition, we develop a content-based image retrieval system that
uses the HIT of a sketched image to retrieve similar images from the
image database. An index structure, basic shape description, is
proposed to implement the candidate selection. It is efficient to
quicken the retrieval time and have high accuracy on selecting
candidates. Experimental results show that the application using these
methods is feasible and has high accuracy on retrieving images.
|