ABSTRACT
Image registration is important in many applications such as computer vision and medical imaging. For medical applications, it is used in clinical studies of disease and for atlas-based identification and segmentation of anatomical structures. The registration process consists in deforming a template image T to match an unbiased reference image R through a smooth invertible transformation. The transformation may further provide information such as the changes of location and degree of deformation to help with diagnosis. For images of the same modality, a well-registered template has geometric features and intensity distribution matched with those of the reference. For images produced by different mechanisms and possessing distinct modalities, the goal of registration is to correlate the images while maintaining the modality of the template.
