| Summary: | 碩士 === 國立中正大學 === 資訊工程學系 === 85 === With the growth of multimedia computing, the amount of images
people used is getting larger and larger. In order to retrieve
the desired imagefrom a large database, we need a correct and
fast method. The methods thathave been proposed use the features
from the spatial domain as the keys forthe image retrieval.
However, in order to save the space for storing images,many
compression methods are used for image compression. Hence, we
can not find the features from compressed images directly. In
this thesis, we proposed a method that uses the edge information
as the keys for image retrieval. Thismethod is used for the JPEG
compressed images. We get the edge information directly from the
compression domain as the keys. Without decompression the
compressed image, we can speed up image retrieval. We detect
the edge information using the relationship between the
DiscreteCosine Transform (DCT) coefficients. All the edge blocks
are then classified into different edge types according to the
edge orientations. The edge typesare used for our image
retrieval. We provide two kinds of queries for users.The first
one uses a partial image from the database images as the query
image.The second query let users sketch the objects of a image
and use this sketchimage as the query image. According to the
experiment results, we know that our method is useful for the
image retrieval.
|
|---|
