Wavelet-Based Image Registration Techniques: A Study of Performance
Nagham E. Mekky, Sherif E. Kishk · 2011
Summary This paper presents a comparative study of performance for four wavelet-based multiresolution image registration techniques. The proposed algorithms are implemented and applied to dental panoramic X-ray images and magnetic resonance (MR) images of the brain. Cross-correlation based registration, mutualinformation (MI) based hierarchical registration, scale invariant feature transform (SIFT) based registration, and hybrid registration approach using MI and SIFT operator combined with wavelet-based hierarchical pyramid, have been utilized. A comparison between proposed techniques with the corresponding techniques in the spatial domain is achieved. The quality of the registration process was measured using the following criteria: normalized cross-correlation coefficient (NCCC) and percentage relative root mean square error (PRRMSE). The application of the selected techniques to dental panoramic X-ray images and brain MR images has shown that wavelet-based hierarchical approach combining MI, SIFT, and RANdom Sample And Consensus (RANSAC) algorithm gives the best results and can be used efficiently for registration of two types of images.