System and method for identifying objects of interest in image data

• US 20060013464A1
• Filed: 05/25/2005
• Published: 01/19/2006
• Est. Priority Date: 05/26/2004
• Status: Active Grant

• Current Assignee: Applied Visual Sciences, Inc.
• Original Assignee: Guardian Technologies LLC (Lasko Holdings, Inc.)
• Inventors: Ramsay, Thomas E., Ramsay, Eugene B., Hamilton, Victor, Richard, Martin, Felteau, Gerard
• Granted Patent: US 7,492,937 B2
• Time in Patent Office: Days
• US Class Current: 382/132
• CPC Class Codes: G06V 10/20 Image preprocessing
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes

Abstract

A system and method for identifying objects of interest in image data is provided. The present invention utilizes principles of dynamic discontinuity in which objects in images, when subjected to special transformations, will exhibit radically different responses based on the chemical properties of the imaged objects. Using the system and methods of the present invention, certain objects that appear indistinguishable from other objects to the eye or computer recognition systems, or are otherwise statistically identical, generate radically different and statistically significant differences that can be easily measured.

Claims

1. A method of identifying a medical object of interest in image data, comprising:

receiving the image data; and

applying at least one bifurcation transform to the image data to effect divergence of the medical object of interest from other objects.

2. The method of claim 1, wherein the at least one bifurcation transform comprises a series of bifurcation transforms.

3. The method of claim 1, wherein the at least one bifurcation transform comprises at least one point operation.

4. The method of claim 3, wherein the at least one point operation is at least partially a non-linear point operation.

5. The method of claim 3, wherein the at least one point operation comprises at least one nodal point.

6. The method of claim 5, wherein the at least one nodal point is adjusted so as to effect the divergence of the medical object of interest from other objects.

7. The method of claim 1, wherein the at least one bifurcation transform is adapted to maintain an integrity of the medical object of interest during divergence of the medical object of interest from other objects.

8. The method of claim 1, further comprising generating a medical output image based on the results of the at least one bifurcation transform.

9. The method of claim 8, wherein the medical object of interest is distinguished from other objects in the output image by adjusting a visual parameter of the medical object of interest and other objects based on the results of the at least one bifurcation transform.

10. The method of claim 1, wherein the image data comprises nonparametric image data.

11. The method of claim 1, wherein the image data comprises parametric image data.

12. The method of claim 1, wherein the image data comprises positron emission tomography (PET) image data.

13. The method of claim 1, wherein the image data comprises ultrasound image data.

14. The method of claim 1, wherein the image data comprises magnetic resonance imaging (MRI) data.

15. The method of claim 1, wherein the image data comprises CT image data.

16. The method of claim 1, wherein the image data comprises x-ray image data.

17. The method of claim 1, wherein the medical object of interest is statistically indistinguishable from other objects in the image data.

18. The method of claim 1, wherein the medical object of interest comprises a plurality of types of medical objects, wherein each type of medical object comprises at least one member.

19. The method of claim 18, wherein the at least one bifurcation transform is generated for each member of the medical objects.

20. The method of claim 18, wherein the type of medical objects include cancers, tissue, organs, cells, bones or abnormalities.

21. The method of claim 1, wherein the image data is scanner data.

22. The method of claim 1, wherein the image data includes a portion of a medical scan.

23. The method of claim 22, wherein the image data includes multiple scans of an input object.

24. The method of claim 1, wherein the image data comprises results from an MRI device, a CAT device, an ultrasound device or an imaging device.

25. An apparatus configured to identify a medical object of interest in image data, comprising:

an input device configured to receive the image data; and

an image transformation recognition system configured to apply at least one bifurcation transform to the image data to effect divergence of the medical object of interest from other objects.

26. A method of creating a bifurcation transform for a class of medical objects, comprising:

selecting a point operation;

performing said point operation on a subset of images, wherein said subset of images includes at least one image containing an object in said class of medical objects; and

repeating said selecting and said performing steps until said point operation bifurcates said object.

27. The method of claim 26, wherein said point operation bifurcates said object by increasing differentials in response characteristics among objects in said class of medical objects relative to other objects in said subset of images.

28. The method of claim 26, wherein said class of medical objects comprises a plurality of types of medical objects, wherein each type of medical object comprises at least one member.

29. An image processing method, comprising:

receiving a plurality of determined objects from an original image;

applying a plurality of response-based image transformations to original image pixels for each of said determined objects to generate a prescribed divergence; and

classifying said determined objects by different response characteristics to at least one of said transforms.

30. The method of claim 29, wherein said plurality of response-based image transforms provide image object bifurcation among the determined objects.

31. The method of claim 29, wherein said plurality of response-based image transforms comprise point transformation functions having at least one nodal point to increase differentials in said response characteristics between said determined image objects.

32. The method of claim 31, wherein said increased differential in said response characteristics determine said image object bifurcation, and wherein said response characteristics comprise pixel color or pixel luminance.

33. The method of claim 32, wherein said increased differentials provide statistically significant color content for pixels of said determined objects.

34. The method of claim 31, wherein said increased differentials occur between a two nodal points in a point transformation function and are approximated by nonlinear functions.

35. The method of claim 31, wherein said point transformation functions comprise grayscale transformation functions having a first characteristic axis having 2^N input levels and a transformed characteristic axis having 2^N output levels, and wherein said at least one nodal point increase said differentials in said response characteristics determined by said output levels of said transformed characteristic axis.

36. The method of claim 29, wherein said plurality of response-based image transforms translate said determined objects into a response-based coordinate domain for comparison.

37. The method of claim 36, wherein said comparison in said response-based coordinate domain comprises:

generating a plurality of identifying feature vectors; and

comparing feature vectors for said determined objects to said identifying feature vectors in said response-based coordinate domain.

38. The method of claim 29, wherein said plurality of response-based image transforms are selectively applied, sequentially applied or concurrently applied, wherein for each determined object resulting transformed color content generated by each image transform is analyzed to determined when criteria of a selected object of interest is satisfied.

39. The method of claim 29, wherein said determined objects comprise image objects generated by satellite surveillance.

40. The method of claim 29, wherein said determined objects comprise image objects generated by aerial surveillance.

41. The method of claim 29, wherein said determined objects comprise image objects generated by X-rays.

42. The method of claim 29, wherein said determined objects comprise image objects generated by at least one sensor detecting a plurality of frequencies.

43. The method of claim 29, wherein said determined objects comprise one of medical image objects and astronomical image objects.

44. The method of claim 29, wherein said determined objects comprise image objects representing land formations.

45. The method of claim 29, wherein said determined objects comprise image objects representing vegetation.

46. The method of claim 29, wherein said determined objects comprise image objects representing man-made objects.

47. The method of claim 29, wherein said classifying said determined objects comprises determining at least one feature object of interest.

48. The method of claim 29, wherein said classifying said determined objects comprises determining at least one area of interest.

49. The method of claim 29, further comprising:

receiving at least one image;

segmenting a first image to identify a subset of objects; and

determining objects images from said segmented objects.

51. The method of claim 29, wherein said plurality of response-based image transforms comprise linear and non-linear point operations, and wherein the images are one of at least of two and three-dimensional representations using pixels or voxels.

52. The method of claim 29, wherein said response-based transformations operate where image discontinuities occur in said original image or discontinuities occur in said determined objects to discriminate objects of said original image.

50. The method of 49, wherein the segmenting comprises:

applying at least one transformation to select objects matching selected color, density or weight characteristics; and

isolating and identifying said object from said image.