Techniques for tracking: Image Registration

Abstract

The process of finding a suitable deformation, s, in order to align two or more images is a process known as image registration, image fusion, matching, or warping. Image registration is the keystone in a myriad of image analysis tasks, especially important in those methods that involve the combination of various data sources, like in image fusion (functional and structural information), change detection (motion, velocity, and deformation), and multichannel image restoration.Registration is also being increasingly used in healthcare to estimate parameters for diagnosis purposes. For example, heart regional motion analysis has proved to be of major importance for the study of cardiac abnormal behavior. It currently plays an indisputable role in treatment and diagnosis of different cardiac pathologies, such as mitral regurgitation [1, 2], ischemia [3], dyssynchrony [4], myocardial quantification [5, 6, 7, 8] and diastolic dysfunction [9]. The analysis of the motion is usually done with registration techniques. There are different imaging modalities used to estimate different regional motion features, dominant among them those based on ultrasound (US) acquisitions. Indeed, echocardiography has become a widely used tool for motion and strain estimation due to low cost and real time acquisition. This chapter is focused on those image registration techniques that estimate themotion and strain by tracking the speckle pattern directly from the intensity of the BmodeUS images. Thus, we leave aside the registration of multimodal images or thegeneration of wide-view US images from acquisitions with different fields of view.Besides, we assume a temporal relationship between images.