ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper F-SIFT and B-RANSAC - New Image Matching Algorithms
Cited - time in scopus Share share facebook twitter linkedin kakaostory
Authors
Jae Kwang Lee, Chang-Joon Park, In-Ho Lee
Issue Date
2007-01
Citation
Korea-Japan Joint Workshop on Frontiers of Computer Vision (FCV) 2007, pp.268-274
Language
English
Type
Conference Paper
Abstract
This paper addresses two image matching algorithms based on SIFT and RANSAC. The proposed algorithms are F-SIFT (Frequency Scale Invariant Feature Transform) and B-RANSAC (Block Random Sample Consensus). F-SIFT is a SIFT algorithm operated in frequency domain not in spatial domain to escape using convolution operator for DoG and the rectangular kernel is used instead of Gaussian for feature extraction. B-RANSAC is an algorithm to avoid repetition and inequality of the selected sample in an image while extracted features are matched. The computational time is slightly reduced by using F-SIFT and the performance of image matching is increased by using B-RANSAC. These algorithms could bring a result of fast and reliable image matching. A series of simulation has been performed to verify the performance for image matching.
KSP Keywords
Computational time, Feature extractioN, Image Matching, Random sample consensus, SIFT Algorithm, frequency domain(FD), matching algorithm, scale invariant feature transform(SIFT), spatial domain