| Summary: | 碩士 === 朝陽科技大學 === 資訊工程系碩士班 === 96 === With the development of computer science and World Wide Web (WWW), the users no longer to satisfy with pure text type information. Diversified multimedia type messages are used in breadth. For this reason, various digital image database is established and growth in magnitude day by day. The traditional text retrieval cannot deal with the requirement of difference data type. Consequently, an efficient content-based image retrieval system is more and more important.
In this thesis, we proposed an appropriate segmentation method, called self-clustering algorithm (SCA), in a color-based retrieval system. The proposed clustering method can make the information of main object more distinct and more suitable for user’s demand. In most retrieval systems, the parameters are always fixed. In order to handle increasing and various images, the Taguchi method is employed for adjusting parameters automatically to achieve higher performance.
Experimental results show that the proposed method can obtain better precision in remaining the same recall. That is to say the proposed system can provide higher performance than the original one that without self-clustering algorithm and Taguchi method.
|
|---|
