Remote Sensing Optical Image Registration Using Modified Uniform Robust SIFT
Uniformly distributed feature extraction and availability of a sufficient number of correctly matched points between the input images are the key challenges for remote sensing optical image registration. Because of its robustness and distinctiveness, the scale-invariant feature transform (SIFT) is a...
Saved in:
Published in | IEEE geoscience and remote sensing letters Vol. 13; no. 9; pp. 1300 - 1304 |
---|---|
Main Authors | , |
Format | Journal Article |
Language | English |
Published |
Piscataway
IEEE
01.09.2016
The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Uniformly distributed feature extraction and availability of a sufficient number of correctly matched points between the input images are the key challenges for remote sensing optical image registration. Because of its robustness and distinctiveness, the scale-invariant feature transform (SIFT) is a well-known approach for an automatic image registration. However, the features obtained from the SIFT algorithm are not uniformly distributed, and sometimes, the number of matched features is insufficient to provide subpixel accuracy in the registration of remote sensing optical images. In this letter, a modified version of SIFT is proposed to obtain uniformly distributed matched features. Then, the bivariate histogram and the random sample consensus have been used to refine the initial matches. Finally, a reliable matching criterion is proposed to maximize the number of matches. |
---|---|
Bibliography: | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23 |
ISSN: | 1545-598X 1558-0571 |
DOI: | 10.1109/LGRS.2016.2582528 |