Summary: | 碩士 === 國立臺灣大學 === 資訊工程學系研究所 === 86 === Digital image databases are becoming more and more popular. Finding images
in large databases becomes a problem. In the past, due to the nature of
image, we usually just can associate some data, such as the title, or
keywords, with an image, and query images by these metadata. This is not a
perfect way. Because the quality of these metadata heavily depend on the
creator, querying by content of the image is a better approach and an
important topic of research.
In this thesis, we propose a model of content based image retrieval
system by our novel hierarchical color region segmentation. We retain the
hierarchical relationship of the regions in an image during segmentation.
Using the information of the relationships and features of the regions, we
can represent the desired objects in images more accurately. In retrieval,
we compare not onlyregion features but also region relationships.
|