Print segmentation system and method

• US 20040218790A1
• Filed: 04/29/2004
• Published: 11/04/2004
• Est. Priority Date: 05/02/2003
• Status: Abandoned Application

• Current Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Ping Lo, Peter Zhen
• Application Number:
  US10/834,536
• Publication Number:
  US 20040218790A1
• Time in Patent Office: Days
• US Class Current: 382/124
• CPC Class Codes: G06V 10/507 Summing image-intensity values; Histogram projection analysis
• CPC Class Codes: G06V 40/1347 Preprocessing; Feature extraction

Abstract

A method for segmenting a fingerprint image having a plurality of pixel includes the steps of: estimating the direction (400) of at least a portion of the plurality of pixels to generate a directional image from said fingerprint image; determining (500) at least one foreground component image of the directional image; calculating (800) a vertical projection profile of the at least one foreground component image; calculating (800) a horizontal projection profile of the at least one foreground component image; determining (1000) a plurality of segmentation edges based on the vertical and horizontal projection profiles; and segmenting (1100) a region of the fingerprint image based on the plurality of segmentation edges, whereby the region is included within the segmentation edges.

Claims

1. A method for segmenting a fingerprint image having a plurality of pixel, said fingerprint image being the image of a fingerprint, said method comprising the steps of:

estimating the direction of at least a portion of said plurality of pixels to generate a directional image from said fingerprint image;

determining at least one foreground component image of said directional image;

calculating a vertical projection profile of said at least one foreground component image;

calculating a horizontal projection profile of said at least one foreground component image;

determining a plurality of segmentation edges based on said vertical and horizontal projection profiles; and

segmenting a region of said fingerprint image based on said plurality of segmentation edges, whereby said region is included within said segmentation edges.

2. The method of claim 1 further comprising preprocessing said fingerprint image before generating said directional image, said preprocessing including down-scaling said fingerprint image.

3. The method of claim 1 further comprising the step of calculating statistical information from said fingerprint image.

4. The method of claim 3, wherein said statistical information includes at least one of a histogram of said fingerprint image, a dynamic range and a local mean value.

5. The method of claim 3, wherein said statistical information is used to modify said directional image.

6. The method of claim 1, said fingerprint image further comprising cores and deltas, and said method further comprising the step of detecting the location of at least a portion of said cores and deltas from said directional image.

7. The method of claim 6, wherein said cores and deltas are used to modify said segmentation edges.

8. The method of claim 1, wherein said at least one foreground component is determined by binarizing said directional image into a two level image comprising a first level image that includes at least one foreground component image having a direction and a second level image comprising at least one background component image having no direction.

9. The method of claim 1 further comprising sorting said at least one foreground component image according to component image size.

10. The method of claim 9 further comprising selecting a largest foreground component image for calculating said vertical and horizontal projection profiles.

11. The method of claim 1 further comprising the step of detecting boundaries of said at least one foreground component image, including at least one left and one right boundary.

12. The method of claim 11 further comprising convexing the boundaries of said at least one foreground component.

13. The method of claim 12 further comprising convexing said at least one foreground component image by computing the convex hull.

14. The method of claim 11 further comprising generating a central line between said at least one left and right boundary for estimating an orientation of said at least one component.

15. The method of claim 14 further comprising rotating said at least one foreground component image to within a predefined threshold of a desired orientation.

16. The method of claim 1 further comprising smoothing said directional image using a multi-cell pyramid approach.

17. The method of claim 1, wherein said fingerprint image is a grayscale image and said plurality of pixels each have a gray value determined from a set of gray values, and said method further comprising down-scaling said set of gray values.

18. The method of claim 17, wherein said set of gray values is down-scaled by a factor of four.

19. The method of claim 1 further comprising:

determining a maximum vertical peak length from said vertical projection profile; and

determining a maximum horizontal peak length from said horizontal projection profile, whereby said plurality of segmentation edges are determined based on said maximum vertical peak length and said maximum horizontal peak length.

20. The method of claim 19 further comprising:

for said vertical projection profile detecting at least one peak rise and a corresponding peak fall based upon a first threshold for determining said maximum vertical peak length; and

for said horizontal projection profile detecting at least one peak rise and a corresponding peak fall based upon a second threshold for determining said maximum horizontal peak length.

21. The method of claim 20, wherein said first and second thresholds are dynamic thresholds.

22. The method of claim 19, wherein said maximum vertical peak length is used to determine a top and bottom segmentation edge, and said maximum horizontal peak length is used to determine a left and right segmentation edge.

23. The method of claim 1, wherein each said segmentation edge has a predetermined length.

24. The method of claim 1 further comprising detecting a fingerprint crease, wherein said region is located above the crease.

25. A system for segmenting a fingerprint image having a plurality of pixel, said fingerprint image being the image of a fingerprint, said system comprising:

means for estimating the direction of at least a portion of said plurality of pixels to generate a directional image from said fingerprint image;

means for determining at least one foreground component image of said directional image;

means for calculating a vertical projection profile of said at least one foreground component image;

means for calculating a horizontal projection profile of said at least one foreground component image;

means for determining a plurality of segmentation edges based on said vertical and horizontal projection profiles; and

means for segmenting a region of said fingerprint image based on said plurality of segmentation edges, whereby said region is included within said segmentation edges.