AUTOMATIC IMAGE REGISTRATION OF MULTI-ANGLE IMAGERY FOR CHRIS/PROBA

Subpixel image registration is the key to successful multi-angle remote sensing image applications such as image fusion, superresolution and classification. However, multi-angle remote sensing images pose some difficulties for automatic image registration, namely, 1) precisely locating control points (CPs) is difficult as large view angle images are susceptible to resolution change and blurring; and 2) local geometric distortion caused by variations in platform stability makes rigid transform models such as the projective model unreliable. In this paper we propose a two-stage automatic registration scheme for multi-angle remote sensing imagery. In the first step, CPs are gathered via the scale invariant feature transform (SIFT). However, CPs collected by SIFT may be too few or unevenly distributed. Therefore, another CPs collecting procedure based on normalized cross-correlation follows. In order to eliminate outliers in the CPs a geometric constraint is utilized; after outlier elimination in order to get CPs of high accuracy for the estimation of the thin plate spline model, which is used to solve the local geometric distortion problem, a pre-fitting procedure is adopted. The methodology developed in this paper is applied to three Compact High Resolution Imaging Spectrometer onboard the Project for On-board Autonomy (CHRIS/Proba) images. Experimental results demonstrate the efficiency and accuracy of the proposed method