Intensity based image registration of satellite images using evolutionary techniques

Abstract

Image registration is the fundamental image processing technique to determine geometrical transformation that gives the most accurate match between reference and floating images. Its main aim is to align two images. Satellite images to be fused for numerous applications must be registered before use. The main challenges in satellite image registration are finding out the optimum transformation parameters. Here in this work the non-alignment parameters are considered to be rigid and affine transformation. An intensity based satellite image registration technique is being used to register the floating image to the native co-ordinate system where the normalized mutual information (NMI) is taken as the similarity metric for optimizing and updating transform parameters. Because of no assumptions are made regarding the nature of the relationship between the image intensities in both modalities NMI is very general and powerful and can be applied automatically without prior segmentation on a large variety of data and as well works better for overlapped images as compared to mutual information(MI). In order to get maximum accuracy of registration the NMI is optimized using Genetic algorithm, particle swarm optimization and hybrid GA-PSO. The random initialization and computational complexity makes GA oppressive, whereas weak local search ability with a premature convergence is the main drawback of PSO. Hybrid GA-PSO makes a trade-off between the local and global search in order to achieve a better balance between convergence speed and computational complexity. The above registration algorithm is being validated with several satellite data sets. The hybrid GA-PSO outperforms in terms of optimized NMI value and percentage of mis-registration error.