政大學術集成


Please use this identifier to cite or link to this item: https://ah.nccu.edu.tw/handle/140.119/74469


Title: Improved sift algorithm to match points in the texture region image
Authors: Chen, Chen C.-Y.;Chio, Shih-Hong
邱式鴻
Contributors: 地政學系
Keywords: Feature point extraction;Harris corner detector;Matching problems;Scale and rotation;Scale invariant feature transforms;SIFT;SIFT algorithms;SIFT descriptors;Algorithms;Feature extraction;Image texture;Remote sensing;Textures;Image matching
Date: 2012
Issue Date: 2015-04-10 16:34:31 (UTC+8)
Abstract: Feature point extraction automatically instead of manually is important and can improve the efficiency for photogrammetric tasks. Scale invariant feature transform (SIFT) is an important algorithm developed by Lowe (2004) to extract keypoints with invariance on scale and rotation for image matching and registration. SIFT uses descriptor to descript a keypoint and employs these descriptors as keypoint's fingerprint in matching. But when it comes to texture region images, SIFT will produce a large number of similar descriptors and these similar descriptors will generatewrong or failedmatching results. In this study, a new algorithm based on SIFT is designed to deal with the matching problem in texture region images. The new algorithm uses the concept of Harris Corner Detector and the entropy in order to make SIFT descriptors more independence. Thus even keypoints are in texture region image, the descriptor of every keypoints is still unique enough to other keypoints. From the tests, this developed new algorithm can make the successful rate of matching up to about 80% in texture region images. The performance of developed algorithm is better than SIFT algorithm.
Relation: 33rd Asian Conference on Remote Sensing 2012, ACRS 2012
Data Type: conference
Appears in Collections:[Department of Land Economics ] Proceedings

Files in This Item:

File Description SizeFormat
index.html0KbHTML799View/Open


All items in 學術集成 are protected by copyright, with all rights reserved.


社群 sharing