Intensity-based 3D local image registration

Abstract

In many imaging applications, including image fusion and image sequence monitoring, we need to geometrically match up one image to another of a same scene, so that information from different images can be compared or combined properly. This is the image registration (IR) problem that has received much attention in recent years due to its broad applications. In the literature, early IR methods are for analyzing 2D images. Because of the rapid progress in image acquisition technologies, 3D images have become increasingly popular in magnetic resonance imaging (MRI) and other applications in recent years. Since the structure of a typical 3D image is substantially more complicated than that of a typical 2D image, 3D image registration is challenging. In this paper, we develop a new 3D image registration method using local smoothing statistical methods. By the flexibility of local smoothing, our method does not require any parametric form or other global regularity conditions on the related geometric transformation. It is shown that this method works well in practice.