Non-rigid image registration using distance functions

• US 20030235337A1
• Filed: 01/31/2003
• Published: 12/25/2003
• Est. Priority Date: 02/01/2002
• Status: Active Grant

• Current Assignee: Siemens Medical Solutions USA Incorporated (Siemens AG)
• Original Assignee: Siemens Medical Solutions USA Incorporated (Siemens AG)
• Inventors: Ramesh, Visvanathan, Rousson, Mikael, Paragios, Nikolaos
• Granted Patent: US 7,200,269 B2
• Time in Patent Office: Days
• US Class Current: 382/215
• CPC Class Codes: G06T 2207/10012 Stereo images
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/149 involving deformable models, e.g. active contour models
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

A system and method for non-rigid image registration using distance functions includes portions for receiving a source shape into an image space, integrating a global linear registration model with local deformations to assess the source shape, optimizing a functional defined on a parameter or feature space where the functional quantifies the similarity between the source shape and a target shape in terms of distance functions, creating an augmented registration space including a plurality of target shape clones coherently positioned in the image space, tracking moving interfaces between the source shape and the target shapes and clones through a level set method, and registering the source shape by seeking mutual correspondences between the source shape, the target shape and the target shape clones.

Claims

1. A method for non-rigid image registration using distance functions, the method comprising:

receiving a plurality of pixels defining a source shape and a target shape into an n-dimensional image space; and

computing a distance map responsive to the shape by calculating a minimum distance from each point in the image space to a closest portion of the shape and defining a higher-dimensional n+1 information space for shape registration.

2. A method as defined in claim 1, further comprising:

cloning the shape within the image space; and

computing a distance map responsive to the cloned shape by calculating a minimum distance from each point in the image space to a closest portion of the cloned shape.

3. A method as defined in claim 2, further comprising performing a global registration that is invariant to rotation, translation and scale, by minimizing an energy cost function in the information space.

4. A method as defined in claim 3 wherein the energy function is optimized to recover global registration parameters for the source shape.

5. A method as defined in claim 3, further comprising minimizing the energy cost function using a gradient-descent method.

6. A method as defined in claim 5, further comprising estimating local deformations from the global registration to optimize the fit of a source shape to the target shape.

7. A method as defined in claim 3 wherein the energy function is responsive to local deformations.

8. A method as defined in claim 7 wherein the global transformation and the local deformations are combined using a local deformation file.

9. A method as defined in claim 8 wherein the local deformation file meets a regular and smooth regularity constraint.

10. A method as defined in claim 9, further comprising minimizing the energy cost function using a gradient-descent method.

11. A method as defined in claim 5, further comprising registering the source shape to a shape model.

12. A method as defined in 11 wherein a global-to-local registration is optimized in accordance with the local quality of the model.

13. A method as defined in claim 12, further comprising minimizing a stochastic cost function in accordance with the quality of the model.

14. A method as defined in claim 13, further comprising minimizing the energy cost function using a gradient-descent method.

15. A method for non-rigid image registration using distance functions, the method comprising:

receiving a source shape and a target shape;

creating an augmented registration space comprising a plurality of target shape clones coherently positioned in an information space;

integrating at least one global linear registration model with local deformations to assess the source shape;

optimizing a functional defined on a parameter or feature space, the functional quantifying the similarity between the source shape and at least one target shape in terms of distance functions; and

registering the source shape by seeking mutual correspondences between the source shape, the target shape and the target shape clones.

16. A method as defined in claim 15 wherein optimizing comprises selecting a mathematical framework by finding an optimum of at least one functional to quantify the similarity between the source shape and at least one target shape.

17. A method as defined in claim 16 wherein the mathematical framework is defined by at least one of variational and stochastic principles.

18. A method as defined in claim 15 wherein optimizing comprises the use of a gradient-descent method within the mathematical framework.

19. A method as defined in claim 15 wherein the image space is greater than a two-dimensional image plane.

20. A method as defined in claim 15 wherein the source shape comprises uncertainties, the method further comprising:

considering a linear parametric model and a pixel-wise deformation field resulting in a non-rigid registration paradigm; and

recovering both of the linear parametric model and a pixel-wise deformation field simultaneously through a gradient-descent method.

21. An apparatus for non-rigid image registration using distance functions, the apparatus comprising:

a receiving unit for receiving a source shape and a target shape;

a cloning unit in signal communication with the optimizing unit for creating an augmented registration space comprising a plurality of target shape clones coherently positioned in the information space;

an integrating unit in signal communication with the receiving unit for integrating at least one global linear registration model with local deformations to assess the source shape;

an optimizing unit in signal communication with the integrating unit for optimizing a functional defined on a parameter or feature space, the functional quantifying the similarity between the source shape and at least one target shape in terms of distance functions; and

a registration unit in signal communication with the tracking unit for registering the source shape by seeking mutual correspondences between the source shape, the target shape and the target shape clones.

22. A system for non-rigid image registration using distance functions, the system comprising:

receiving means for receiving a source shape and a target shape;

cloning means in signal communication with the optimizing means for creating an augmented registration space comprising a plurality of target shape clones coherently positioned in the information space;

integrating means in signal communication with the receiving means for integrating at least one global linear registration model with local deformations to assess the source shape;

optimizing means in signal communication with the integrating means for optimizing a functional defined on a parameter or feature space, the functional quantifying the similarity between the source shape and at least one target shape in terms of distance functions; and

registration means in signal communication with the tracking means for registering the source shape by seeking mutual correspondences between the source shape, the target shape and the target shape clones.

23. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform program steps for non-rigid image registration using distance functions, the program steps comprising:

receiving a source shape and a target shape;

creating an augmented registration space comprising a plurality of target shape clones coherently positioned in the information space;

integrating at least one global linear registration model with local deformations to assess the source shape;

optimizing a functional defined on a parameter or feature space, the functional quantifying the similarity between the source shape and at least one target shape in terms of distance functions; and

registering the source shape by seeking mutual correspondences between the source shape, the target shape and the target shape clones.