Method and System for Edge Detection

• US 20140112541A1
• Filed: 12/31/2013
• Published: 04/24/2014
• Est. Priority Date: 03/24/2008
• Status: Active Grant

• Current Assignee: Cognyte Technologies Israel Ltd. (Cognyte Software Ltd.)
• Original Assignee: Verint Systems Limited (Verint Systems Incorporated)
• Inventors: Golan, Oren, Kiro, Shmuel, Horovitz, Itshak
• Granted Patent: US 8,965,054 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06T 5/00 Image enhancement or restoration
• CPC Class Codes: G06T 7/13 Edge detection
• CPC Class Codes: G06V 10/431 Frequency domain transformation; Autocorrelation
• CPC Class Codes: G06V 10/443 by matching or filtering
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

A method executed by a computer system for detecting edges comprises receiving an image comprising a plurality of pixels, determining a phase congruency value for a pixel, where the phase congruency value comprises a plurality of phase congruency components, and determining if the phase congruency value satisfies a phase congruency criteria. If the phase congruency value satisfies the phase congruency criteria, the computer system categorizes the pixel as an edge pixel. If the phase congruency value does not satisfy the phase congruency criteria, the computer system compares a first phase congruency component of the plurality of phase congruency components to a phase congruency component criteria. If the first phase congruency component satisfies the phase congruency component criteria, the computer system categorizes the pixel as an edge pixel, and if the first phase congruency component does not satisfy the phase congruency component criteria, categorizes the pixel as a non-edge pixel.

Claims

1-25. -25. (canceled)

26. A method of operating a computer system for edge detection, the method comprising:

receiving a plurality of images, each image of the plurality comprising a plurality of pixels;

determining a phase congruency value for each pixel of the plurality based upon at least a first phase congruency component;

producing a list of edge pixels from the determined phase congruency values;

producing, a list of corner pixels from the determined phase congruency values;

producing an edge image from the list of edge pixels and list of corner pixels using non-maximal suppression;

producing an enhanced edge image by processing the edge image using hysteresis; and

applying an object detection algorithm to the enhanced edge image to identify at least one object in the plurality of images.

27. The method of claim 26, wherein the non-maximal suppression comprises:

identifying local maxima pixels between identified corner pixels as selected edge pixels; and

discarding all other edge pixels, wherein the edge image is constructed from the selected edge pixels.

28. The method of claim 26, wherein using hysteresis comprises:

comparing an energy of the pixels in an edge direction which are not in the edge image to a hysteresis threshold;

identifying those pixels with energy greater than the hysteresis threshold as new edge pixels; and

adding the new edge pixels to the edge image to produce the enhanced edge image.

29. The method of claim 26, further comprising detecting objects within the plurality of images from the categorized edge pixels.

30. The method of claim 29, wherein the detected objects are people within the plurality of images and further comprising:

tracking a quantity, location, or movement of the detected objects.

31. The method of claim 30, further comprising counting people entering, leaving, or remaining within a defined space.

32. The method of claim 29, wherein the detected objects are portions of people.

33. The method of claim 32, wherein the detected objects are heads of people within the plurality of images.

34. The method of claim 26, further comprising:

determining if the phase congruency values satisfies a first criteria;

if the phase congruency value of a pixel satisfies the first criteria, adding the pixel to the list of edge pixels or the list of corner pixels;

if the phase congruency value of the pixel does not satisfy the first criteria, determining if the first phase congruency component satisfies a second criteria different from the first criteria; and

if the first phase congruency component of the pixel satisfies the second criteria adding pixel to the list of edge pixels or the list of corner pixels.

35. The method of claim 34, further comprising:

processing the received plurality of images using at least one filter to produce first intermediate images for two or more scales and orientations.

processing the first intermediate images to determine noise energy;

producing second intermediate images comprising the determined noise energy; and

processing the first intermediate images and the second intermediate images to produce third intermediate images comprising local energy compensated for noise.

36. The method of claim 35, wherein the at least one filter is a plurality of logarithmic Gabor filters.

37. The method of claim 35, further comprising, processing the first intermediate images and the third intermediate images to produce phase congruency data for each of the received images.

38. The method of claim 37, further comprising:

computing maximal moments of phase congruency for each orientation from the phase congruency data;

comparing the maximal moments of phase congruency for each orientation to a phase congruency threshold; and

based upon the comparison, producing the list of edge pixels.

39. The method of claim 37, further comprising:

computing minimal moments of phase congruency for each orientation from the phase congruency data;

determining axes wherein the minimal moments of phase congruency are minimized; and

based upon the determination, producing the list of corner pixels.

40. A method of operating a computer system for edge detection, the method comprising:

receiving a plurality of images, each image of the plurality comprising a plurality of pixels;

determining a phase congruency value for each pixel of the plurality based upon at least a first phase congruency component;

producing, a list of edge pixels from the determined phase congruency values;

producing a list of corner pixels from the determined phase congruency values;

identifying local maxima pixels between identified corner pixels as selected edge pixels;

producing an edge image from the selected edge pixels;

comparing an energy of pixels in an edge direction which are not in the edge image to a hysteresis threshold;

identifying those pixels with energy greater than the hysteresis threshold as new edge pixels;

producing an enhanced edge image by adding the new edge pixels to the edge image; and

applying an object detection algorithm to the enhanced edge image to identify at least one object in the plurality of images.

41. The method of claim 40, further comprising detecting objects within the plurality of images from the categorized edge pixels.

42. The method of claim 41, wherein the detected objects are people within the plurality of images and further comprising tracking a quantity, location, or movement of the detected objects.

43. The method of claim 40, comprising:

wherein determining the phase congruency value comprises computing maximal moments of phase congruency for each pixel;

comparing the maximal moments to a phase congruency threshold to produce the list of edge pixels;

wherein determining the phase congruency value further comprises computing minimal moments of phase congruency for each pixel; and

determining axes wherein minimal moments of phase congruency are minimized to produce the list of corner pixels.

44. A non-transitory computer-readable medium having instructions stored thereon for operating a computer system to detect edges, wherein the instructions, when executed by the computer system, direct the computer system to:

receive a plurality of images, each image of the plurality comprising a plurality of pixels;

determine a phase congruency value for each pixel of the plurality based upon at least a first phase congruency component;

produce a list of edge pixels from the determined phase congruency values;

produce a list of corner pixels from the determined phase congruency values;

produce an edge image from the list of edge pixels and list of corner pixels using non-maximal suppression;

produce an enhanced edge image by processing the edge image using hysteresis; and

apply an object detection algorithm to the enhanced edge image to identify at least one object in the plurality of images.

45. The non-transitory computer-readable medium of claim 44, further comprising:

wherein the non-maximal suppression comprises identifying local maxima pixels between identified corner pixels as selected edge pixels and discarding all other edge pixels;

wherein the edge image is constructed from the selected edge pixels; and

wherein using hysteresis comprises comparing an energy of the pixels in an edge direction which are not in the edge image to a hysteresis threshold, identifying those pixels with energy greater than the hysteresis threshold as new edge pixels, and adding the new edge pixels to the edge image to produce the enhanced edge image.