Study on eliminating wrong match pairs of SIFT

To eliminate or reduce wrong pairs in SIFT(Scale Invariant Feature Transform)feature matching, this paper proposes an error canceling algorithm through computing match-base. First finding three match point pairs by considering minimal Euclidean distance; second take three match point pairs as match-...

Full description

Saved in:
Bibliographic Details
Main Authors: Ren Jiayuan, Wang Yigang, Deng Yun
Format: Conference Proceeding
Language:English
Subjects:
Blood
Complexity theory
eliminating worng match
Euclidean distance
Feature extraction
Graphics
image mosaic
Nearest neighbor searches
RANSAC
Robustness
SIFT
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To eliminate or reduce wrong pairs in SIFT(Scale Invariant Feature Transform)feature matching, this paper proposes an error canceling algorithm through computing match-base. First finding three match point pairs by considering minimal Euclidean distance; second take three match point pairs as match-base if these pairs satisfy with criterion, or finding another three match pairs. Finally checking all other match pairs and eliminating wrong match pairs. Experimental results show that the proposed algorithm consumes lower time than RANSAC with inheriting rotation invariance of SIFT. At the same time, it does not exclude the right match pairs, with strong robustness. As a result, precision of image mosaic has been improved.
ISSN:2164-5221
DOI:10.1109/ICOSP.2010.5655847