Chapter 3: Segmentation

Title: Features correspondence of remote sensing images

Author(s): Bo Tang, Djamel Ait-Boudaoud, Lik-Kwan Shark, Bogdan J. Matuszewski

Address: ADSIP Research Centre, University of Central Lancashire, Preston PR1 2HE, UK | ADSIP Research Centre, University of Central Lancashire, Preston PR1 2HE, UK | ADSIP Research Centre, University of Central Lancashire, Preston PR1 2HE, UK | ADSIP Research Centre, University of Central Lancashire, Preston PR1 2HE, UK

Reference: Atlantic Europe Conference on Remote Imaging and Spectroscopy pp. 65 - 71

Abstract/Summary: In this paper the features correspondence problem is investigated by using three types of features extracted by the Harris corner, Harris affine corner and SIFT (Scale Invariant Feature Transform) detectors. Using three matching strategies adapted to each point feature detection method to estimate the features correspondence, the performances of the three feature detection methods are evaluated using remote sensing image pairs. The results show that the Harris corner is the most computation efficient method, but only works well for short baseline stereo image pairs. The Harris affine and the SIFT methods are able to cope with both of the short and wide baseline image pairs. However their computation efficiencies are low, especially for the Harris affine method.

Order a copy of this article