A dual metric approach towards similarity measurement in image registration

Image registration is an essential pre-processing step in a wide range of image processing applications such as medical imaging, terrain recognition from aerial images, biometric identification procedures etc. Recognition, matching or comparison of images can be done once the images that we seek to identify are registered. A crucial part of the image registration technique is to measure the degree of similarity between the two images, usually referred to as the test image and reference image, respectively. Several metrics have been used in the literature to measure this degree of similarity. In this paper, we compare the performance of the correlation coefficient(CC) and mutual information (MI) in terms of their sensitivity to translation and rotation in intra-modal image registration. We then propose the use of a dual metric consisting of both CC and MI to achieve efficient and accurate image registration. We demonstrate our results using fingerprint images obtained from the NIST database and iris images.