Confirmed boundary pattern matching

• US 5,027,422 A
• Filed: 08/29/1988
• Issued: 06/25/1991
• Est. Priority Date: 08/29/1988
• Status: Expired due to Term

• Current Assignee: Raytheon Company (Rtx Corporation)
• Original Assignee: Raytheon Company (Rtx Corporation)
• Inventors: Kanter, Irving, Peregrim, Theodore J., Kettering, Gordon L.
• Primary Examiner(s): Tarcza, Thomas H.
• Assistant Examiner(s): CAIN, DAVID C
• Application Number:
  US07/237,927
• Time in Patent Office: 1,030 Days
• Field of Search: 382/22, 382/34, 382/41, 382/48, 382/49, 382/50, 382/27, 382/56, 364/135
• US Class Current: 382/294
• CPC Class Codes: G06T 5/80 Geometric correction
• CPC Class Codes: G06V 10/36 Applying a local operator, i.e. means to operate on image points situated in the vicinity of a given point; Non-linear local filtering operations, e.g. median filtering
• CPC Class Codes: G06V 10/757 Matching configurations of points or features

Abstract

A method of aligning two images of the same scene by matching features in a first image to features in a second image is disclosed. The method comprises identifying edges of objects in the first image using two different processes. The edges identified using both processes are compared and combined into one image representing confirmed edges which are readily identified in other images of the same scene. A template is then formed from the confirmed edges which is matched to a subregion of the second image.

Claims

1. A method of aligning two images of the same scene, each of such images comprising an array of pixels, said method comprising the steps of:

(a) forming a template map of a portion of a first image, the template map comprising an array of pixels, selected ones of said pixels having;

(i) a first value to indicate that the corresponding pixels in the first image represents a portion of the image, said portion being near an edge of the image and having an intensity greater than the intensity of another portion of the image on another side of the edge;

or

(ii) a second value to indicate that the corresponding pixel in the first image represents a portion of the image, said portion being near an edge of the image and having an intensity less than the intensity of another portion on another side of the edge;

(b) forming a correlation threshold indicative of the ratio of the average intensity of pixels in the template map having the first value to the average intensity of pixels in the first image corresponding to pixels in the template map having the second value;

(c) applying the template map to a plurality of local windows of pixels formed in the second image, computing a score for each local window, and selecting the local window producing the highest score to represent alignment, and wherein computing a score for each local window comprises the steps of;

(i) computing a first threshold equal to the inverse of the correlation threshold multiplied by the average value of pixels in the local window of the second image corresponding to pixels in the template map having the first value;

(ii) computing a second threshold equal to the correlation threshold multiplied by the average value of pixels in the local window of the second image corresponding to pixels in the template map having the second value;

(iii) computing a score reflecting the number of pixels in the local window having values above the second threshold and corresponding to pixels in the template map having the first value plus the number of pixels in the local window having values below the first threshold and having the corresponding to pixels in the template map second value pluse the number of pixels in the local window having values above the second threshold and corresponding to pixels in the template map having the second value, plus the number of pixels in the local window having values below the first threshold and corresponding to pixels in the template map having the first value.

2. The method of aligning two images as in claim 1 wherein the score is computed according to the formula

3. The method of aligning two images as in claim 2 wherein:

(a) the first constant is between 0 and 2; and

(b) the second constant is between 1 and 7.