AISIR: Automated inter-sensor/inter-band satellite image registration using robust complex wavelet feature representations

Abstract

An automated registration system named AISIR (Automated inter-sensor/inter-band satellite image registration) has been designed and implemented for the purpose of registering satellite images acquired using different sensors and spectral bands. Sensor and environmental noise, contrast non-uniformities, and inter-sensor and inter-band intensity mapping differences are addressed in the AISIR system. First, a novel modified Geman–McClure M-estimation scheme using a robust phase-adaptive complex wavelet feature representation is introduced for robust control point matching. Second, an iterative refinement scheme is introduced in the AISIR system for improved control point pair localization. Finally, the AISIR system introduces a robust mapping function estimation scheme based on the proposed modified Geman–McClure M-estimation scheme. The AISIR system was tested using various multi-spectral optical, LIDAR, and SAR images and was shown to achieve better registration accuracy than state-of-the-art M-SSD and ARRSI registration algorithms for all of the test sets.