Image registration using closest corresponding voxels with an iterative registration process
First Claim
1. A method of registering three-dimensional (3D) volume images, wherein a volume image is represented by an array of voxel color values characterizing physical parameters of a volume of a 3D object, wherein the 3D object is characterized, at least in part, by the volume images being registered, the method comprising the steps of:
- capturing a first volume image, the first volume image characterizing a first set of physical characteristics of a scanned 3D object;
capturing a second volume image, the second volume image characterizing a second set of physical characteristics of the scanned 3D object;
generating voxel pairs, wherein each voxel pair comprises a first voxel from a first of the two volume images and a second voxel from the second of the two volume images, when the second voxel has a characteristic that is common to both the first voxel and the second voxel and the second voxel is at least as close to the first voxel as any other voxel in the second volume image having the characteristic; and
iteratively determining a registration transformation from the generated voxel pairs, where the registration transformation is a transformation relating the two volume images.
1 Assignment
0 Petitions
Accused Products
Abstract
A robust, automatic volume registration process based on intensity gradients successfully performs registrations under conditions of unrelated intervolume voxel intensities, significant object displacements and/or significant amounts of missing data. The process allows a user to visualize the registration convergence, clearly illustrating any source of registration errors. The process includes steps of matching, based on iteratively finding a correspondence of the closest voxels containing a high three-dimensional (3D) intensity gradient magnitude. The process is a powerful method of sequence-independent MR volume registration which is simple to both use and understand.
-
Citations
19 Claims
-
1. A method of registering three-dimensional (3D) volume images, wherein a volume image is represented by an array of voxel color values characterizing physical parameters of a volume of a 3D object, wherein the 3D object is characterized, at least in part, by the volume images being registered, the method comprising the steps of:
-
capturing a first volume image, the first volume image characterizing a first set of physical characteristics of a scanned 3D object; capturing a second volume image, the second volume image characterizing a second set of physical characteristics of the scanned 3D object; generating voxel pairs, wherein each voxel pair comprises a first voxel from a first of the two volume images and a second voxel from the second of the two volume images, when the second voxel has a characteristic that is common to both the first voxel and the second voxel and the second voxel is at least as close to the first voxel as any other voxel in the second volume image having the characteristic; and iteratively determining a registration transformation from the generated voxel pairs, where the registration transformation is a transformation relating the two volume images. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 19)
-
-
17. A method of registering two-dimensional (2D) images, wherein an image is represented by an array of pixel color values characterizing physical parameters of an area of an object, wherein the object is characterized, at least in part, by the images being registered, the method comprising the steps of:
-
capturing a first image characterizing a first set of physical characteristics of a scanned object; capturing a second image characterizing a second set of physical characteristics of the scanned object; generating pixel pairs wherein each pixel pair comprises a first pixel from a first of the two images and a second pixel from the second of the two images, when the second pixel has a characteristic that is common to both the first pixel and the second pixel and the second pixel is at least as close to the first pixel as any other pixel in the second image having the characteristic; and iteratively determining a registration transformation from the generated pixel pairs, where the registration transformation is a transformation relating the two images. - View Dependent Claims (18)
-
Specification