A Region-Based Image Indexing and Retrieval method
碩士 === 國立中正大學 === 電機工程研究所 === 89 === The objective of this thesis is to build an effective image retrieval system to help people manage a large number of image informations . Our system is based on the structure of the region-based image retrieval method . Modified watershed algorithm is...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2001
|
Online Access: | http://ndltd.ncl.edu.tw/handle/93629085222641687426 |
Summary: | 碩士 === 國立中正大學 === 電機工程研究所 === 89 === The objective of this thesis is to build an effective image retrieval system to help people manage a large number of image informations . Our system is based on the structure of the region-based image retrieval method . Modified watershed algorithm is applied for image segmentation . Color features are choosen to describe the ROI ( region of interesting ) .The K-mean algorithm is then used to cluster feature vectors of ROI to build the database . After the image database has been built example images can be used as input for image retrieval . The query system use the same segmentation and feature extraction process as the database built phase . then MSE algorithm is used to find the nearest class . The similarity test algorithm is proposed to calculate the similarity of the images . The result is sorted by similarity and output to the users .
This study has the following contribution:
(1)Use image objects as the main body of the database .
(2)Use a modified watershed algorithm to meet the segmentation of the region-based approach .
(3)Emphasize the importance of color features in region-based image retrieval .
(4)A new algorithm is proposed to calculate the similarity between objects .
|
---|