Machine print, hand print, and signature discrimination

• US 9,940,511 B2
• Filed: 05/29/2015
• Issued: 04/10/2018
• Est. Priority Date: 05/30/2014
• Status: Active Grant

• Current Assignee: Kofax Incorporated
• Original Assignee: Kofax Incorporated
• Inventors: Shustorovich, Alexander, Thrasher, Christopher W., Macciola, Anthony, Amtrup, Jan W.
• Primary Examiner(s): Kholdebarin, Iman K
• Assistant Examiner(s): TRAN, MAI H
• Application Number:
  US14/726,335
• Publication Number:
  US 20150347836A1
• Time in Patent Office: 1,047 Days
• CPC Class Codes: G06V 10/273 removing elements interfering with the pattern to be recognised
• CPC Class Codes: G06V 30/36 Matching; Classification
• CPC Class Codes: G06V 40/382 Preprocessing; Feature extraction

Abstract

Systems, computer program products, and techniques for discriminating hand and machine print from each other, and from signatures, are disclosed and include determining a color depth of an image, the color depth corresponding to at least one of grayscale, bi-tonal and color; reducing color depth of non-bi-tonal images to generate a bi-tonal representation of the image; identifying a set of one or more graphical line candidates in either the bi-tonal image or the bi-tonal representation, the graphical line candidates including one or more of true graphical lines and false positives; discriminating any of the true graphical lines from any of the false positives; removing the true graphical lines from the bi-tonal image or the bi-tonal representation without removing the false positives to generate a component map comprising connected components and excluding graphical lines; and identifying one or more of the connected components in the component map.

Claims

1. A method, comprising:

receiving an image;

determining a color depth of the received image, the color depth corresponding to at least one of grayscale, bi-tonal and color;

reducing color depth of the received image to generate a bi-tonal representation of the image in response to determining the color depth of the received image is not bi-tonal;

identifying a set of one or more graphical line candidates in either the bi-tonal image or the bi-tonal representation, the graphical line candidates including one or more of true graphical lines and false positives;

discriminating any of the true graphical lines from any of the false positives;

removing the true graphical lines from the bi-tonal image or the bi-tonal representation without removing the false positives to generate a component map comprising connected components and excluding graphical lines;

identifying one or more of the connected components in the component map; and

one or more of;

outputting an indicator of each of the connected components, each indicator identifying one or more of the connected component and a particular class to which the connected component corresponds; and

displaying the indicator of each of the connected components; and

wherein the indicator comprises one or more of a label identifying the particular class to which the connected component corresponds, and a bounding box surrounding the connected component.

2. The method as recited in claim 1, the identifying further comprising:

determining a connectedness of the one or more connected components in the component map;

determining whether each of the one or more connected components corresponds to a particular class of components based on the connectedness thereof, wherein the class of components to which each of the one or more connected components corresponds is selected from unknown, handwritten characters, signature, and machine-printed characters; and

labeling each of the one or more connected components based on the particular class of components to which the connected component corresponds.

3. The method as recited in claim 2, wherein determining the connectedness of the one or more connected components comprises determining, for each of the one or more connected components, a number of connected component transitions across a vertical threshold position.

7. The method as recited in claim 3, further comprising defining the vertical threshold position at approximately a midpoint of a maximum vertical stroke height of each of the one or more connected components prior to determining the connectedness of the connected component.

8. The method as recited in claim 3, further comprising defining the vertical threshold position at approximately two thirds of a maximum vertical stroke height of each of the one or more connected components prior to determining the connectedness of the connected component.

9. The method as recited in claim 3, further comprising defining the vertical threshold position at approximately three quarters of a maximum vertical stroke height of each of the one or more connected components prior to determining the connectedness of the connected component.

10. The method as recited in claim 2, further comprising one or more of outputting and displaying an indicator of each connected component, each indicator identifying one or more of the connected component and the particular class to which the connected component corresponds, and

wherein the indicator comprises one or more of the label identifying the particular class to which the connected component corresponds, and a bounding box surrounding the connected component.

11. The method as recited in claim 10, the outputting comprising associating each indicator as metadata with one or more of the bi-tonal image, the bi-tonal representation, and the received image, and

wherein the metadata identify at least a position of the connected component and the particular class to which the connected component corresponds.

12. The method as recited in claim 11, wherein the position of the connected component identifies each pixel of the image forming the connected component.

13. The method as recited in claim 2, further comprising performing a neighbor analysis to determine whether the particular class to which one or more adjacent components is identical to the particular class to which the connected component corresponds;

and either

confirming the particular class to which the connected component corresponds in response to determining the particular class to which the one or more adjacent components corresponds is identical to the particular class to which the connected component corresponds;

or

refuting a label associated with the connected component and identifying the particular class to which the connected component corresponds in response to determining the particular class to which the connected component corresponds is different than the particular class to which the one or more adjacent components correspond.

14. The method as recited in claim 13, wherein the method comprises the refuting, and the neighbor analysis further comprising, in response to refuting the connected component label, modifying the particular class to which the connected component corresponds to match the particular class to which one or more of the adjacent connected components correspond.

15. The method as recited in claim 1, wherein determining whether each connected component corresponds to a particular class of connected components comprises one or more of support vector analysis, maximum entropy discrimination analysis, and neural network analysis.

16. The method as recited in claim 1, wherein the discriminating comprises filtering the image to remove one or more at least partially globular features from the set of graphical line candidates;

and

wherein the filtering is based on one or more of feature length, feature thickness, feature pixel density, and feature size.

17. The method as recited in claim 16, wherein the set of graphical line candidates are characterized by a thickness less than a threshold thickness, the thickness being measured in a direction substantially perpendicular to an orientation along which the true graphical line lies.

18. The method as recited in claim 17, wherein the threshold thickness corresponds to a value less than about 0.1 inches.

4. A method, comprising:

receiving an image;

determining a color depth of the received image, the color depth corresponding to at least one of grayscale, bi-tonal and color;

reducing color depth of the received image to generate a bi-tonal representation of the image in response to determining the color depth of the received image is not bi-tonal;

identifying a set of one or more graphical line candidates in either the bi-tonal image or the bi-tonal representation, the graphical line candidates including one or more of true graphical lines and false positives;

discriminating any of the true graphical lines from any of the false positives;

removing the true graphical lines from the bi-tonal image or the bi-tonal representation without removing the false positives to generate a component map comprising connected components and excluding graphical lines;

identifying one or more of the connected components in the component map, the identifying further comprising;

determining a connectedness of the one or more connected components in the component map, wherein determining the connectedness of the one or more connected components comprises determining, for each of the one or more connected components, a number of connected component transitions across a vertical threshold position;

determining whether each of the one or more connected components corresponds to a particular class of components based on the connectedness thereof, wherein the class of components to which each of the one or more connected components corresponds is selected from unknown, handwritten characters, signature, and machine-printed characters, wherein the particular class to which the connected component corresponds is determined to be a signature in response to determining the number of connected component transitions is at least 5; and

labeling each of the one or more connected components based on the particular class of components to which the connected component corresponds.

5. The method as recited in claim 4, wherein the particular class to which the connected component corresponds is determined to be a signature further in response to determining the connected component is characterized by a maximum stroke height less than about 70% of a width of the connected component.

6. The method as recited in claim 4, wherein the particular class to which the connected component corresponds is determined to be a signature further in response to determining the connected component is characterized by a width of at least 0.4 inches.

19. A computer program product comprising:

a non-transitory computer readable storage medium having program code embodied therewith, the program code readable and/or executable by a processor to;

receive an image;

determine a color depth of the received image, the color depth corresponding to at least one of grayscale, bi-tonal and color;

reduce color depth of the received image to generate a bi-tonal representation of the image in response to determining the color depth of the received image is not bi-tonal;

identify a set of one or more graphical line candidates in the bi-tonal representation, the graphical line candidates including one or more of true graphical lines and false positives;

discriminate any of the true graphical lines from any of the false positives;

remove the true graphical lines from the bi-tonal representation without removing the false positives to generate a component map comprising connected components and excluding graphical lines;

identify one or more of the connected components in the component map, wherein identifying the one or more of the connected components in the component map comprises;

determining a connectedness of the one or more connected components in the component map;

determining whether each of the one or more connected components corresponds to a particular class of components based on the connectedness thereof, wherein the class of components to which each of the one or more connected components corresponds is selected from unknown, handwritten characters, signature, and machine-printed characters; and

labeling each of the one or more connected components based on the particular class of components to which the connected component corresponds; and

one or more of outputting and displaying an indicator of each connected component, each indicator identifying one or more of the connected component and the particular class to which the connected component corresponds; and

wherein each indicator respectively comprises one or more of the label identifying the particular class to which the connected component corresponds, and a bounding box surrounding the connected component.