Automatic registration of high resolution satellite images using artificial immune system and mutual information

摘要

Image registration is the process that allows geometric alignment of two images by determining the transformation that provides the most accurate match between two images. It is a crucial underlying process in many remote sensing applications such as multi-temporal classification, change detection, environmental monitoring, cloud removal, video geo-registration and map updating, etc. In this paper, we introduce a new approach for automated image registration without the need of sensors parameters and control points. The main characteristic of the algorithm is the use of a powerful search strategy based on artificial immune system (AIS) and mutual information as similarity measure. The method was implemented and tested using a variety of high-resolution satellite imagery such as Ikonos (0.8 metre resolution), QuickBird (0.8 m) and WorldView-2 (0.5 metres) respectively taken from Brazil, Iran and China. Experimental results and comparative studies demonstrate the effectiveness of the proposed approach for registration of high resolution satellite images.