By Topic

Pixel-Based Invariant Feature Extraction and its Application to Radiometric Co-Registration for Multi-Temporal High-Resolution Satellite Imagery

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
Yonghong Li ; Department of Electrical and Computer Engineering, University of Missouri-Columbia, Columbia, MO, USA ; Curt H. Davis

Here we present a robust fully automated method for relative radiometric co-registration. First, a new low dimensional feature-point descriptor, called the Expanded Haar-Like Filter (EHLF) descriptor, is introduced. The EHLF has many desirable properties like flexible design, fast computation, and multi-scale description, while also being insensitive to variations in image quality. Next, two spatial matching schemes are proposed for increasing the percentage of correctly matched feature points. The first is based on a global affine model and the second utilizes dynamic local template fuzzy distance matching. Finally, precise pixel-to-pixel invariant feature points are extracted from a diversity of image locations centered at matched local extrema points. Experimental results show that for high-resolution multi-temporal imagery, the EHLF descriptor can obtain matched feature points with accuracies equivalent to that using a higher dimensional descriptor. In addition, the EHLF descriptor produces a larger number of correctly matched feature points. The spatial matching methods significantly improve feature-point matching, especially for image pairs with large geometric distortions. Radiometric co-registration quality based on the pixel-based invariant features was tested using four different evaluation datasets, and the results demonstrate that the proposed approach produces the lowest normalized root-mean-square error compared with six other automated methods. The proposed method depends on successful extraction of feature points, which may not be available for the scenes that are fully undeveloped (e.g., forest areas). Nonetheless, Monte Carlo simulations show that 30 to 50 correctly matched feature points will provide relatively stable radiometric calibration coefficients.

Published in:

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing  (Volume:4 ,  Issue: 2 )