Image registration under translation using a multiresolution hybrid metric approach
Anagha Shah, P. Mithun, Vikram M. Gadre · 2009
Registration in medical imaging has applications in radiation treatment planning, image based navigation during surgery, motion correction and fusion of anatomical and functional information. A similarity metric is crucial for correct registration of images. Mutual information is a widely used similarity metric for registration of multimodality and monomodality medical images. However, it lacks spatial information and hence can lead to misregistration. A hybrid metric composed of mutual information and spatial information is suggested in the literature which gives better result. We have applied a wavelet-based multiresolution scheme to 2D sets of moving and fixed images where in mutual information and spatial information metrics are calculated from the approximation and sub bands respectively. This approach makes the registration faster. We have validated the better performance of the hybrid metric, calculated from the aforementioned approach, for image registration when the moving image is translated over the fixed image along x and y directions. (5 pages)