Object extraction in images

• US 5,923,776 A
• Filed: 05/23/1996
• Issued: 07/13/1999
• Est. Priority Date: 05/23/1996
• Status: Expired due to Fees

• Current Assignee: the united states of america as represented by the secretary of the navy
• Original Assignee: the united states of america as represented by the secretary of the navy
• Inventors: Kamgar-Parsi, Behrooz
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US08/652,044
• Time in Patent Office: 1,146 Days
• Field of Search: 382/205, 382/270, 382/101, 382/102, 382/308, 382/173, 358/455, 358/465, 358/464, 358/453
• US Class Current: 382/173
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns

Abstract

A method and apparatus for extracting an object from an image, in which one locates a pixel within the image (the "central" pixel), and then sequentially compares the brightness of neighboring pixels, proceeding outward from the central pixel. In so doing, one determines the largest dropoffs in brightness between neighboring pixels, and uses these to determine a brightness threshold for extracting pixels belonging to the object. In a preferred embodiment, one determines the threshold by comparing the largest dropoffs, identifies overlapping regions of brightness level common to all the dropoffs, and sets the threshold at the midpoint of the common overlapping region.

Claims

1. A method for extracting an object from an image, comprising:

forming an image containing an object, said method further comprising;

a. disposing a pair of transverse axes on said object;

b. choosing a pixel on one of said axes, said pixel being denominated the reference pixel;

c. choosing a neighboring pixel, said neighboring pixel being located away from said reference pixel in a direction transverse to said one of said axes;

d. determining the dropoff in brightness between said neighboring pixel and said reference pixel;

e. replacing said reference pixel by said neighboring pixel;

f. repeating steps c through e a preselected number of times;

g. selecting a preselected number L, and selecting the L largest of said dropoffs; and

h. using said largest of said dropoffs to determine a threshold for extracting said object.

3. The method of claim 1, wherein said choosing of said reference pixel comprises:

selecting a preselected number of pixels in said object, said preselected number of pixels being the brightest of said image; and

using said preselected number of pixels for determining said reference pixel.

4. The method of claim 3, wherein said determining of said reference pixel comprises averaging the spatial position of said preselected number of pixels within said image.

5. The method of claim 4, wherein said averaging is the averaged spatial position of said preselected number of pixels within said image, weighted by the intensities of said preselected number of pixels.

6. The method of claim 3, wherein said selecting a preselected number of pixels further comprises:

a'"'"'. determining whether any of said preselected number of pixels are more distant from the others of said preselected number of pixels by a preselected nearness criterion;

b'"'"'. eliminating from said preselected number of pixels those which are more distant by said preselected nearness criterion; and

repeating said steps a'"'"' and b'"'"' until the number of pixels in said preselected number of pixels becomes constant.

7. The method of claim 1, wherein said step h to determine said threshold for extracting said object comprises:

determining the overlapping range of said dropoffs; and

setting said threshold to be within said range.

8. The method of claim 7, wherein said setting of said threshold causes said threshold to be the midpoint of said range.

9. The method of claim 1, wherein said step g for selecting said L largest of said dropoffs causes eliminating from said largest of said dropoffs those which exceed a preselected magnitude.

10. The method of claim 1, wherein said method further comprises extracting said object by:

comparing each pixel in said image to said image to said threshold; and

identifying as part of said object each pixel whose intensity exceeds said threshold.

11. The method of claim 1, wherein said method further comprises:

a". identifying pixels immediately adjacent said reference pixel having eight-fold symmetry with said reference pixel;

b". comparing each of said pixels immediately adjacent said reference pixel to said threshold to identify those pixels which exceed said threshold; and

c". identifying as part of said object said those pixels which exceed said threshold.

12. The method of claim 11, wherein said those pixels which exceed said threshold are denominated the reference pixel set, said method further comprising:

a'"'"'". for each member of said pixel set;

b'"'"'". identifying neighboring pixels to said each member, said neighboring pixels being immediately adjacent said each member and having eight-fold symmetry therewith;

c'"'"'". determining which of said each of said neighboring pixels exceeds said threshold;

d'"'"'". identifying as part of said object said those neighboring pixels which exceed said threshold;

e'"'"'". making those pixels identified in step d'"'"'" a new reference pixel set; and

f'"'"'". iteratively repeating steps a'"'"'" through e'"'"'" until the number of pixels identified as part of said object becomes constant.

13. The method of claim 1, wherein said choosing said reference pixel comprises:

selecting a preselected number of the brightest pixels in said image;

applying a selected nearness criterion to each of said preselected number of said pixels to determine if said each of said preselected number of pixels is distant from the others of said preselected number by a preselected amount;

removing from said preselected number a pre set number of the least bright pixels; and

iteratively repeating said steps for applying and removing, until each of said preselected number of pixels is distant from the others of said preselected number by said preselected amount.

2. An apparatus for extracting an object from an image, said image containing an object, comprising:

an imager for forming said image, and a processor, said processor being adapted for;

disposing of a pair of transverse axes on said object;

choosing a pixel on one of said axes, said pixel being denominated the reference pixel;

choosing a neighboring pixel, said neighboring pixel being located away from said reference pixel in a direction transverse to said one of said axes;

determining the dropoff in brightness between said neighboring pixel and said reference pixel;

replacing said reference pixel by said neighboring pixel;

repeating said choosing a pixel, choosing a neighboring pixel, determining said dropoff, and said replacing said reference pixel, said repeating being done a preselected number of times;

said apparatus further comprising;

means for selecting a preselected number L, and selecting the L largest of said dropoffs; and

means for using said largest of said dropoffs to determine a threshold for extracting said object.