Mutual Information Based Image Registration: Approach Using Wavelet Decomposition

Abstract

Image registration is m apping positions in one image with the corresponding positions in the other image.
Rigid registration o f images using hierarchical decomposition is useful because a coarse to fin e strategy
helps in correctly estimating the mapping function parameters as we move to a finer resolution at each
level. This also reduces search space and computational time. Multi resolution nature o f wavelets is
employed for image decomposition. The decom posed images are then registered using maximization o f
Mutual Information (MI). In this paper variation o f MI and signal to noise ratio (SNR) with increase in
level o f decomposition and with use o f Sub-sampling/filtering an d wavelet decomposition has been
analyzed. Results with wavelet decomposition have been fou n d to be better than those with subsampling! filtering as with each reduced search space finer details are being analyzed rather than the entire
image content and this is the inherent feature ofw avelet decomposition.