Multimodal Medical Image registration using Discrete Wavelet Transform

Abstract

In image processing, image registration develops a relationship between two images using optimal transformation where the images could have been taken at various times, sources or devices, or from different perspectives. It aligns the reference and moving image using geometric transformations. This research study evaluates the performance of multimodal(images acquired from different sources)image registration technique using Discrete Wavelet Transform (DWT).The reference and the target images are decomposed into their respective DWT coefficients and then are processed for image registration. After registration, the resultant DWT coefficients are transformed back using Inverse DWT into their spatial coordinates in order to retrieve the registered image. The similarity of the two input images for image registration is calculated and investigated using a similarity metric known as Mutual Information (MI) which is maximized. The quality of registration is measured using cross-correlation coefficient (CCC) of the registered image with respect to the reference image. Finally the time taken for image registration in wavelet domain is analyzed and compared with the image registration taking place in spatial domain.