| Summary: | 碩士 === 國立臺灣大學 === 電機工程學系 === 86 === Abstract
With the exponential growth of multi-media data comes the challenge of
managing this information in a controlled, organized and efficient manner.
Color, texture and shape are the main features for human beings to recognize
the object. Many researchers use these features to index and
search the images.
We will review the related work on the topic of color image query in chapter 2.
Several systems have been developed to help people to index the database
and search the images .We discuss three reported systems in chapter 3.
Although there were much work on benchmarking text database, the property
of multimedia is different from the text database. A.Desainarasimhalu et al.
analyze the problem of benchmarking multimedia databases and suggest a
methodology. We describe it in Chapter4.
After the study of algorithms, we designed a new system to implement what
we have studied. The procedure could be divided into 2 stages -feature
extraction and query. In feature extraction, system automatically extracts the
content of each image in the database as the index for query. In the query
stage, system matches the features of user's sketch with the features of
database image, and provides a ranked list whose rank is according to the
similarity. We discuss the procedure in chapter 5.
|
|---|
