Effective image retrieval techniques based on novel salient region segmentation and relevance feedback |
| |
Authors: | Min-Yuan Fang Yu-Hsin Kuan Chung-Ming Kuo Chaur-Heh Hsieh |
| |
Affiliation: | (1) Department of Information Engineering, I-Shou University, Tahsu, 840 Kaohsiung, Taiwan, Republic of China;(2) Dept. of Computer and Communication Engineering, Ming Chung University, Gui-Shan, 333 Taoyuan, Taiwan, Republic of China; |
| |
Abstract: | This paper proposes an effective region-based image retrieval technique based on novel salient region segmentation and relevance
feedback. With a good and fast segmentation technique, our system achieves an on-the-fly segmentation capability, which enables
users to select particular regions for matching and feedbacks without waiting for image segmentation. Therefore, we adopt
a relatively simple feedback schemes to derive the intent of the user. The experimental results show that the system performance
is greatly improved with this capability. Furthermore, a Quick-match algorithm is also presented in this paper. The mechanism
of the Quick-match algorithm is to exclude from distance computation regions that are of low possibility to be the top-Mmatches. This algorithm excludes most of regions from distance computation and therefore greatly cuts down the turnaround
time of the retrieval with slightly degradation of precision. |
| |
Keywords: | |
本文献已被 SpringerLink 等数据库收录! |
|