Technique for edge/corner detection/tracking in image frames

• US 5,311,305 A
• Filed: 06/30/1992
• Issued: 05/10/1994
• Est. Priority Date: 06/30/1992
• Status: Expired due to Term

• Current Assignee: American Telephone & Telegraph Company (AT&T, Inc.)
• Original Assignee: AT&T, Inc.
• Inventors: Mahadevan, Ramaswamy, Nalwa, Vishvjit S.
• Primary Examiner(s): Chin, Tommy P.
• Assistant Examiner(s): LEE, RICHARD J
• Application Number:
  US07/906,916
• Time in Patent Office: 679 Days
• Field of Search: 358/105, 358/222, 358/125
• US Class Current: 348/169
• CPC Class Codes: G01S 3/7864 T.V. type tracking systems
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: H04N 19/523 with sub-pixel accuracy
• CPC Class Codes: H04N 19/54 using feature points or meshes
• CPC Class Codes: H04N 7/183 for receiving images from a single remote source

Abstract

Edges/corners in an image can be detected and tracked over time by first sampling the image at periodic time intervals, and then processing the samples to obtain the intensity value for each pixel within each image acquired at a particular time. For each pixel, the bidirectional intraframe correlation of its intensity with the intensities of pixels along each of several directions within the same image is computed. Also, the bidirectional interframe correlation of the intensity of each pixel in each image with the intensities of pixels along each of several directions spanning several images is established. The intraframe and interframe intensity correlations are processed to obtain spatiotemporal tangent information about isobrightness surfaces and curves which are indicative of edges and corners, respectively, and their image motions.

Claims

1. A method for detecting and tracking edges/corners in an image captured by a camera comprising the steps of,

periodically sampling the image at regular time intervals;

processing each sampled image to establish an intensity for each of a plurality of individual picture elements (pixels) collectively comprising said sampled image;

computing an intraframe bidirectional correlation of the intensity of each pixel in each said sampled image with each of the pixels within a predetermined neighborhood in said image to detect line and surface singularities in spatiotemporal space;

computing an interframe bidirectional correlation of the intensity of each pixel in said each sampled image with each of the pixels in a predetermined neighborhood in each of other sampled images to detect line and surface singularities in spatiotemporal space and the movement of said line and surface singularities; and

determining, from said interframe and intraframe bidirectional correlations, the presence of an edge/corner in each sampled image and its motion over time.

2. The method according to claim 1 wherein a corner is determined by:

computing a maximum β

_max from among interframe intensity correlations;

computing a maximum c.sub.⊥

from among interframe intensity correlation components that are orthogonal to a maximum interframe intensity correlation;

computing a maximum α

_max from among intraframe intensity correlations;

comparing both of a pair of ratios c.sub.⊥

/β

_max and α

_max /β

_max to a predetermined threshold; and

establishing the presence of a corner when both ratios c.sub.⊥

/β

_max and α

_max /β

_max are less than a predetermined threshold.

3. The method according to claim 2 wherein a corner is determined by:

comparing said ratios c.sub.⊥

/β

_max and α

_max /β

_max to a predetermined threshold;

computing a maximum e.sub.⊥

from among each interframe and intraframe intensity correlation component that lies along a cross product of a maximum intraframe intensity correlation with a maximum interframe intensity correlation;

comparing ratio e.sub.⊥

/α

_max to a predetermined threshold;

detecting whether there exists an inflection in pixel intensities in a direction orthogonal to a plane defined by α

_max and β

_max ; and

establishing the presence of an edge when;

(a) at least one of the ratios c.sub.⊥

/β

_max and α

_max /β

_max exceeds a predetermined threshold;

(b) the ratio e.sub.⊥

/α

_max is less than a predetermined threshold; and

(c) there is an inflection in pixel intensities in a direction orthogonal to the plane defined by α

_max ×

β

_max.

4. The method according to claim 2 wherein the motion of a corner is established by evolution of β

_max over time.

5. The method according to claim 2 wherein the motion of an edge is established by evolution of (α

_max ×

β

_max) over time.