System and method for detecting an object in a high dimensional space

• US 8,009,900 B2
• Filed: 09/17/2007
• Issued: 08/30/2011
• Est. Priority Date: 09/28/2006
• Status: Active Grant

• Current Assignee: Siemens Healthineers AG (Siemens AG)
• Original Assignee: Siemens Medical Solutions USA Incorporated (Siemens AG)
• Inventors: Yang, Jing, Barbu, Adrian, Georgescu, Bogdan, Zheng, Yefeng, Comaniciu, Dorin
• Primary Examiner(s): CARTER, AARON W
• Application Number:
  US11/856,208
• Publication Number:
  US 20080085050A1
• Time in Patent Office: 1,443 Days
• Field of Search: 382/154, 382/159, 382/160, 382/285, 345419-427, 356 12- 14, 348 42- 60
• US Class Current: 382/154
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data

Abstract

A system and method for detecting an object in a high dimensional image space is disclosed. A three dimensional image of an object is received. A first classifier is trained in the marginal space of the object center location which generates a predetermined number of candidate object center locations. A second classifier is trained to identify potential object center locations and orientations from the predetermined number of candidate object center locations and maintaining a subset of the candidate object center locations. A third classifier is trained to identify potential locations, orientations and scale of the object center from the subset of the candidate object center locations. A single candidate object pose for the object is identified.

Claims

1. A method for detecting an object in a high dimensional image space comprising:

receiving a three dimensional image of an object;

training, by a processor, a first classifier in the marginal space of the object center location which generates a predetermined number of candidate object center locations;

training, by the processor, a second classifier to identify potential object center locations and orientations from the predetermined number of candidate object center locations and maintaining a subset of the candidate object center locations;

training, by the processor, a third classifier to identify potential locations, orientations and scale of the object center from the subset of the candidate object center locations; and

identifying a single candidate object pose for the object.

2. The method of claim 1 wherein the first classifier is trained using 3D Haar features that are subsampled in the image.

3. The method of claim 2 wherein a Probabilistic Boosting Tree is used to train the 3D Haar features.

4. The method of claim 1 wherein the second classifier is trained using curvature features.

5. The method of claim 4 wherein the second classifier works in six dimensional space of locations and orientations.

6. The method of claim 1 wherein the third classifier is trained in nine dimensional space of locations, orientations and scale.

7. The method of claim 1 wherein the object is a left ventricle.

8. The method of claim 1 wherein object is a left atrium.

9. The method of claim 7 wherein the 3D image is a 3D computed tomography image.

10. A system for detecting objects in a high dimensional image space comprising:

an acquisition device for acquiring three dimensional images of an object;

a processor that receives the acquired three dimensional images of the object, the processor performing the following steps on each image;

training a first classifier in the marginal space of the object center location which generates a predetermined number of candidate object center locations,

training a second classifier to identify potential object center locations and orientations from the predetermined number of candidate object center locations and maintaining a subset of the candidate object center locations,

training a third classifier to identify potential locations, orientations and scale of the object center from the subset of the candidate object center locations, and

identifying a single candidate object pose for the object;

and

a display for displaying the detected object.

11. The system of claim 10 wherein the first classifier is trained using 3D Haar features that are subsampled in the image.

12. The system of claim 11 wherein a Probabilistic Boosting Tree is used to train the 3D Haar features.

13. The system of claim 10 wherein the second classifier is trained using curvature features.

14. The system of claim 13 wherein the second classifier works in six dimensional space of locations and orientations.

15. The system of claim 10 wherein the third classifier is trained in nine dimensional space of locations, orientations and scale.

16. The system of claim 10 wherein the object is a left ventricle.

17. The system of claim 10 wherein object is a left atrium.

18. The system of claim 16 wherein the 3D image is a 3D computed tomography image.