Extracting user data from a scanned image of a pre-printed form

• US 6,330,357 B1
• Filed: 04/07/2000
• Issued: 12/11/2001
• Est. Priority Date: 04/07/1999
• Status: Expired due to Term

• Current Assignee: Matthews International Corporation
• Original Assignee: RAF Technology, Incorporated (Constellation Software Incorporated)
• Inventors: Elmenhurst, Brian J., Tyler, Richard H.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): SHERALI, ISHRAT I
• Application Number:
  US09/544,802
• Time in Patent Office: 613 Days
• Field of Search: 382/175-179, 382/180, 382/186-189, 382/199, 382/284, 382/202, 382/305, 382/267, 382/137, 382/162-164, 382/254, 382/173, 382/171, 382/112, 358/501, 358/538, 358/462-464, 358/401, 358/452, 358/453, 707/505-508
• US Class Current: 382/175
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/155 Removing patterns interfering with the pattern to be recognised, such as ruled lines or underlines

Abstract

Human and machine readability of pre-printed forms that have been completed with user data is impeded where the user data overlaps zone descriptions, constraint boxes or other markings of the pre-printed form. A “form fracturing” methodology is described that includes processing the composite-image data so as to attach one or more shared pixels to a non-diagonally adjacent data pixel. The remaining form pixels can be removed, resulting in at least a useful approximation and often a complete recovery of the user data. Where blank-form data is not available, a “virtual dropout” technique allows for recovering user data from a pre-printed form using limited speckle size and configurations, constraining gray-scale value, or a combination of the two. The disclosed methodologies are conveniently implemented in software on any digital processor.

Claims

1. A method for extracting data pixels from a composite image, the composite image consisting of digital data defining a rectilinear array of pixels representative of a pre-printed form having data entered thereon, the method comprising the steps of:

identifying all shared pixels in the composite image;

defining as non-shared data pixels all (black) data pixels that are not common to the pre-printed form;

attaching to the non-shared data pixels all shared pixels that are non-diagonally adjacent to at least one non-shared data pixel; and

removing from the composite image all pixels that are part of the pre-printed form except the attached shared pixels, thereby forming a data image having at least a useful approximation of the data pixels.

2. A method according to claim 1 wherein the data entered on the pre-printed form is hand-written data.

3. A method according to claim 1 wherein the data entered on the pre-printed form is alpha-numeric data.

4. A method according to claim 1 wherein the pre-printed form is a medical claim form.

5. A method according to claim 1 wherein the composite image is formed by scanning.

6. A method according to claim 1 wherein the composite image data is pre-processed so that each pixel has one of two predetermined values.

7. A method according to claim 1 wherein the pre-printed form includes a zone description, whereby the zone description is substantially removed from the composite image.

8. A method according to claim 1 wherein said attaching step includes attaching at least two shared pixels to each non-shared data pixel.

9. A method according to claim 8 wherein the attaching step includes attaching shared pixels having a speckle size of six or less.

10. A method according to claim 8 including selecting the attached pixels configuration based on expected data point size.

11. A computer-implemented method for extracting user data from a composite image, the composite image consisting of digital data defining a matrix of pixels responsive to user data entered on a pre-printed form, or a portion thereof, the method comprising the steps of:

receiving a first matrix F of form pixel data defining the pre-printed form;

receiving a second matrix X of composite image pixel data defining the composite image;

determining an intersection of the F and X matrices to form a third matrix S defining shared pixels;

subtracting the third matrix S from the second matrix X to form a fourth matrix Y defining non-shared user data pixels;

comparing the shared pixels S to the non-shared user data pixels Y to identify all shared pixels that are non-diagonally adjacent to at least one non-shared user data pixel, to form a fifth matrix Z defining attached pixels; and

then

subtracting the pre-printed form pixels F, except for the attached pixels Z, from the composite image pixels X to recover at least a useful approximation of the user data pixels.

12. A method according to claim 11 wherein said comparing the shared pixels S to the non-shared user data pixels Y to form the matrix Z defining attached pixels includes incorporating a plurality of shared pixels into the Z matrix in response to a single, adjacent user data pixel.

13. A method according to claim 11 wherein each such plurality of shared pixels included in the Z matrix defining attached pixels consists of a speckle having a predetermined maximum speckle size.