Image registration using closest corresponding voxels with an iterative registration process

• US 5,937,083 A
• Filed: 04/28/1997
• Issued: 08/10/1999
• Est. Priority Date: 04/29/1996
• Status: Expired due to Fees

• Current Assignee: Government Of The United States Of America, Represented By The Secretary, Department Of Health And Human Services
• Original Assignee: United States Department Of Health And Human Services (Government of the United States of America)
• Inventors: Ostuni, John
• Primary Examiner(s): Kelley, Christopher S
• Assistant Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US08/847,733
• Time in Patent Office: 834 Days
• Field of Search: 382/154, 382/131, 382/285, 382/294, 364/413.13, 364/413.22, 395/119, 395/125, 395/127
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/30016 Brain
• CPC Class Codes: G06T 3/14 Transformations for image registration, e.g. adjusting or mapping for alignment of images
• CPC Class Codes: G06T 7/33 using feature-based methods
• CPC Class Codes: G06T 7/38 Registration of image sequences

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.

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.

2. The method of claim 1, wherein the characteristic that is in common is a characteristic of the first voxel and the second voxel having a common value for a voxel property.

3. The method of claim 1, wherein the characteristic that is in common is a characteristic of the first voxel and the second voxel having similar values for a voxel property.

4. The method of claim 3, wherein the similar values are both values within a predetermined range of values.

5. The method of either claim 2 or claim 3, wherein the voxel property is a function of a voxel intensity.

6. The method of either claim 2 or claim 3, wherein the voxel property is a function of a voxel intensity gradient.

7. The method of either claim 2 or claim 3, wherein the voxel property is a function of a voxel intensity value and neighboring voxel intensity values.

8. The method of either claim 2 or claim 3, wherein the voxel property is a function of a binarized voxel intensity gradient.

9. The method of claim 1, wherein the step of generating voxel pairs excludes voxels meeting a predetermined exclusion criteria.

10. The method of claim 9, wherein the predetermined exclusion criteria is a function of voxel color value.

11. The method of claim 9, wherein the predetermined exclusion criteria is a function of voxel intensity.

12. The method of claim 9, wherein the predetermined exclusion criteria is whether a voxel has an intensity value within an exclusion range.

13. The method of claim 9, wherein the predetermined exclusion criteria is whether a voxel has an intensity value outside an inclusion range.

14. The method of claim 9, wherein the predetermined exclusion criteria is a function of voxel location.

15. The method of claim 9, wherein the predetermined exclusion criteria is whether a voxel is located within an exclusion range.

16. The method of claim 9, wherein the predetermined exclusion criteria is whether a voxel is located outside an inclusion range.

19. The method of claim 1, further comprising the steps of:

generating a second set of voxel pairs, wherein each voxel pair in the second set of voxel pairs comprises a first voxel from the first volume image and a second voxel from the second volume image, when the first voxel has a characteristic that is common to both the first voxel and the second voxel and the first voxel is at least as close to the second voxel as any other voxel in the first volume image having the characteristic; and

wherein the step of iteratively determining a registration transformation is performed using the voxel pairs generated in the step of generating voxel pairs and using the voxel pairs generated in the step of generating the second set of voxel pairs.

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.

18. The method of claim 17, further comprising the steps of:

generating a second set of pixel pairs, wherein each pixel pair in the second set of pixel pairs comprises a first pixel from the first image and a second pixel from the second image, when the first pixel has a characteristic that is common to both the first pixel and the second pixel and the first pixel is at least as close to the second pixel as any other pixel in the first image having the characteristic; and

wherein the step of iteratively determining a registration transformation is performed using the pixel pairs generated in the step of generating pixel pairs and using the pixel pairs generated in the step of generating the second set of pixel pairs.