Automatic template and field definition in form processing

• US 6,886,136 B1
• Filed: 05/05/2000
• Issued: 04/26/2005
• Est. Priority Date: 05/05/2000
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Shevach, Ronen, Kagan, Oren, Wallach, Eugene, Zlotnick, Aviad
• Primary Examiner(s): Bayerl, Raymond J.
• Assistant Examiner(s): BONSHOCK, DENNIS G
• Application Number:
  US09/566,058
• Time in Patent Office: 1,817 Days
• Field of Search: 358/462, 382/317, 715/505, 715/780
• US Class Current: 715/780
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G06T 9/001 Model-based coding, e.g. wire frame
• CPC Class Codes: G06V 30/412 Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables
• CPC Class Codes: G07C 9/37 using biometric data, e.g. fingerprints, iris scans or voice recognition

Abstract

A method for processing a plurality of input images containing variable content that is filled into respective, fixed templates. The method includes comparing the images to collect a group of the images having a high degree of similarity therebetween, and combining the images in the group so as to distinguish the variable content from a fixed portion common to a preponderant number of the images in the group. The fixed portion is processed to reconstruct the fixed template that is common to at least some of the images among the preponderant number, and information is extracted from the images using the reconstructed template.

Claims

1. A method for processing a plurality of input images containing variable content that is filled into respective, fixed templates, the method comprising:

storing a set of one or more reference templates in a memory;

comparing the images to collect a group of the images that do not match any of the one or more stored reference templates and which have a high degree of similarity therebetween;

combining the images in the group so as to distinguish the variable content from a fixed portion common to a preponderant number of the images in the group;

processing the fixed portion to reconstruct a new template, not previously stored in the memory, that is common to at least some of the images among the preponderant number; and

extracting information from the images using the reconstructed template.

2. A method according to claim 1, wherein combining the images comprises mutually aligning and summing the images.

3. A method according to claim 2, wherein each of the input images comprises a plurality of pixels having respective pixel values, and wherein summing the images comprises generating a gray-scale image in which the gray-scale value of each pixel is a sum of the pixel values of the corresponding pixels in the input images.

4. A method according to claim 3, wherein the pixels in the fixed portion of the gray-scale image are characterized generally by higher gray-scale values than the other pixels in the gray-scale image.

5. A method according to claim 3, wherein combining the images comprises binarizing the gray-scale image, so that the pixels in the fixed portion generally receive a first binary value, and the pixels corresponding to the variable content generally receive a second, different binary value.

6. A method according to claim 5, wherein binarizing the gray-scale image comprises testing a plurality of binarization thresholds so as to find one of the threshold that optimally distinguishes the variable content from the fixed portion.

7. A method according to claim 1, wherein processing the fixed portion comprises choosing one of the images from among the preponderant number and combining the chosen image with the fixed portion common to the preponderant number in order to extract the new template from the chosen image.

8. A method according to claim 7, wherein choosing the one of the images comprises choosing an image having a high measure of similarity to the fixed portion.

9. A method according to claim 1, wherein extracting the information comprises removing the reconstructed template from the images.

10. A method according to claim 1, wherein extracting the information comprises finding, responsive to the template, respective locations of fields in the images that contain the information and automatically extracting the information contained in the fields.

11. A method according to claim 10, wherein reading the information contained in the fields comprises:

processing one or more of the images in the group manually, so as to record, for each of the fields, the information filled into the field in conjunction with an identification of the field; and

automatically comparing the manually-recorded information with the automatically-extracted information to find, for each of the fields, a correspondence between the identification of the field and its location.

12. A method according to claim 1, wherein the input images comprise images of documents on which the respective templates are pre-printed, and wherein the variable content comprises alphanumeric characters filled into fields of the templates.

13. A method for processing a group of images containing variable information that is filled into a common template, the method comprising:

processing one or more of the images automatically to determine respective locations of fields in the images that are filled in with the variable information;

automatically extracting the information from the fields;

processing the one or more images manually, so as to record by manual key-in, for each of the fields, the information filled into the field in conjunction with an identification of the field;

automatically comparing the manually-recorded information with the automatically-extracted information to find, for each of the fields, a correspondence between the identification of the field and its location; and

processing additional images in the group to automatically extract the variable information therefrom and, using the correspondence, to identify the extracted information.

14. A method according to claim 13, wherein automatically extracting the information comprises reading alphanumeric characters contained in the fields.

15. A method according to claim 13, wherein processing the additional images comprises detecting, in one of the additional images, information in a new location that does not correspond to any of the fields for which the correspondence was found, and sending the one of the additional images to be processed manually.

16. A method according to claim 15, wherein automatically extracting the information from the fields comprises extracting the information in the new location, and wherein automatically comparing the manually-recorded information with the automatically-extracted information comprises finding a correspondence between the new location and an identification of a field in the new location.

17. A method according to claim 13, wherein the images comprise images of documents on which the template is pre-printed, and wherein the variable information comprises alphanumeric characters filled into fields of the templates.

18. Image processing apparatus, comprising:

an image input device, which is adapted to provide a plurality of input images containing variable content that is filled into respective, fixed templates;

a memory, which is arranged to store a set of one or more reference templates; and

a processor, which is adapted to receive and compare the images to collect a group of the images that do not match any of the one or more stored reference templates and which have a high degree of similarity therebetween, to combine the images in the group so as to distinguish the variable content from a fixed portion common to a preponderant number of the images in the group, to process the fixed portion to reconstruct the fixed a new template, not previously stored in the memory, that is common to at least some of the images among the preponderant number, and to extract information from the images using the reconstructed template.

19. Apparatus according to claim 18, wherein the input images comprise images of documents on which the respective templates are pre-printed, and wherein the variable content comprises alphanumeric characters filled into fields of the templates.

20. Image processing apparatus, comprising:

an image input device, which is adapted to provide a group of images containing variable information that is filled into a common template; and

a processor, which is adapted to process one or more of the images automatically to determine respective locations of fields in the images that are filled in with the variable information, to automatically extract the information from the fields and to compare the automatically-extracted information with manually-extracted information, recorded by manual key-in in manual processing of the images and including, for each of the fields, the information filled into the field in conjunction with an identification of the field, so as to automatically find, for each of the fields, a correspondence between the identification of the field and its location, and to process additional images in the group to automatically extract the variable information therefrom and, using the correspondence, to identify the extracted information.

21. Apparatus according to claim 20, wherein the images comprise images of documents on which the template is pre-printed, and wherein the variable information comprises alphanumeric characters filled into fields of the templates.

22. A computer software product for processing a plurality of input images containing variable content that is filled into respective, fixed templates, wherein a set of one or more reference templates is stored in a memory accessible to a computer, the product comprising a computer-readable medium in which computer program instructions are stored, which instructions, when read by the computer, cause the computer to compare the images to collect a group of the images that do not match any of the one or more stored reference templates and which have a high degree of similarity therebetween, to combine the images in the group so as to distinguish the variable content from a fixed portion common to a preponderant number of the images in the group, to process the fixed portion to reconstruct a new template, not previously stored in the memory, that is common to at least some of the images among the preponderant number, and to extract information from the images using the reconstructed template.

23. A computer software product for processing a group of images containing variable information that is filled into a common template, the product comprising a computer-readable medium in which computer program instructions are stored, which instructions, when read by a computer, cause the computer to process one or more of the images automatically to determine respective locations of fields in the images that are filled in with the variable information, to automatically extract the information from the fields and to compare the automatically-extracted information with manually-extracted information, recorded by manual key-in in manual processing of the images and including, for each of the fields, the information filled into the field in conjunction with an identification of the field, so as to automatically find, for each of the fields, a correspondence between the identification of the field and its location, and to process additional images in the group to automatically extract the variable information therefrom and, using the correspondence, to identify the extracted information.