Method and device for the real-time localization of rectilinear contours in a digitized image, notably for shape recognition in scene analysis processing

• US 5,233,670 A
• Filed: 07/22/1991
• Issued: 08/03/1993
• Est. Priority Date: 07/31/1990
• Status: Expired due to Fees

• Current Assignee: Thomson Trt Defense
• Original Assignee: Thomson Trt Defense
• Inventors: Dufour, Jean-Yves, Le Gall, Serge, Waldburger, Hugues
• Primary Examiner(s): Mancuso, Joseph
• Application Number:
  US07/733,807
• Time in Patent Office: 743 Days
• Field of Search: 382/22, 382/27, 382/54, 382/52, 382/25, 382/41
• US Class Current: 382/197
• CPC Class Codes: G06T 2207/10032 Satellite or aerial image; Remote sensing
• CPC Class Codes: G06T 2207/10048 Infrared image
• CPC Class Codes: G06T 2207/30184 Infrastructure
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06V 10/457 by analysing connectivity, e.g. edge linking, connected component analysis or slices

Abstract

Apparatus and method to localize rectilinear contours in a digitized image for recognizing shapes at a scene. The gradient of the gray level function of the image at the position of each pixel is determined, and those pixels which constitute a contour pixel are identified. The identified contour pixels are complemented with filler pixels where discontinuities are found. The neighborhood of pixels about a contour pixel is compared with a series of characteristic pixel configurations to determine if a correspondence exists with the characteristic configurations. In this way, rectilinear contour pixels are identified.

Claims

1. A method to localize rectilinear contours in a digitized image, for the recognition of shapes in a scene analysis processing operation, said image being formed by a two-dimensional frame of pixels, each exhibiting a determined gray level,

wherein said method comprises the steps of:

(a) approximating for each pixel the gradient of the gray level function of the image at the position of each pixel, said gradient being defined by an argument, representing an element of direction information, and by a norm, representing an element of amplitude information of the transition of the gray level in said direction,

(b) extracting from among all of said pixels a sub-set of contour pixels, where each contour pixel corresponds to a local maximum of the gray level function in the direction of the gradient, the maximum being determined from said amplitude information,

(c) complementing the sub-set of contour pixels by interposed filler pixels if a discontinuity of the corresponding contour exists in a given neighborhood, and

(d) performing for each contour pixel or filler pixel a comparison between the neighborhood of the contour or filler pixel with a series of characteristic pixel configurations, and designating the corresponding pixel as being a rectilinear contour pixel if the comparison determines a correspondence of the neighborhood of the pixel with one of the characteristic configurations.

2. The method of claim 1, wherein said neighborhood being compared includes two concentric and non-contiguous square crowns defined around each contour pixel or filler pixel.

3. The method of claim 2, wherein the two concentric and non-contiguous square crowns are constituted by a 5×

5 crown and a 9×

9 crown surrounding the contour pixel or filler pixel.

4. The method of claim 1, wherein the step (a) for approximating the gradient comprises forming the convolution of the gray level function of the image with the gradient of a smoothing function.

5. The method of claim 4, wherein said smoothing function is a derivative of the Gaussian function.

6. The method of claim 1 further comprising determining whether the norm of the gradient is below a threshold and forcing the value of the norm to zero if the norm is below said threshold.

7. The method of claim 6, wherein said threshold determining operation is an adaptive thresholding operation, wherein respective thresholds are determined as a function of a mean of a norm of a gradient on the previous image.

8. A device for localizing rectilinear contours in a digitized image, for recognizing shapes in a scene analysis processing operation, said image being formed by a two-dimensional frame of pixels each exhibiting a determined gray level, comprising:

means for approximating for each pixel the gradient of the gray level function of the image at the position of each pixel, said gradient being defined by an argument, representing an element of amplitude information and by a norm, representing an element of amplitude information of the transition of the gray level in said direction,

discriminating means for extracting from among said pixels a sub-set of contour pixels, where each contour pixel corresponds to a local maximum of the gray level function in the direction of the gradient, said maximum being determined from transitions of said amplitude information,

filler means for complementing said sub-set of contour pixels by interposing filler pixels if there is a discontinuity in the corresponding contour in a given neighborhood, and

comparator means for comparing for each contour pixel or filler pixel, the neighborhood of the pixel with a series of characteristic configurations, and designating the corresponding pixel as being a rectilinear contour pixel is a comparison determines a correspondence between the neighborhood of the pixel with one of said characteristic configurations.

9. The device of claim 8, wherein localizing of the rectilinear contours is done in real time.