Optical character recognition

• US 10,176,392 B2
• Filed: 01/31/2014
• Issued: 01/08/2019
• Est. Priority Date: 01/31/2014
• Status: Active Grant

• Current Assignee: Micro Focus IP Development Ltd. (Open Text Corporation)
• Original Assignee: Longsand Limited (Open Text Corporation)
• Inventors: Blanchflower, Sean
• Primary Examiner(s): Krasnic, Bernard
• Application Number:
  US15/114,783
• Publication Number:
  US 20160342852A1
• Time in Patent Office: 1,803 Days
• Field of Search: None
• CPC Class Codes: G06F 18/25 Fusion techniques
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06F 18/29 Graphical models, e.g. Bayesian networks
• CPC Class Codes: G06V 30/224 of printed characters having additional code marks or containing code marks

Abstract

Optical character recognition is described in various implementations. In one example implementation, a method may include receiving a plurality of optical character recognition (OCR) outputs provided by a respective plurality of OCR engines, each of the plurality of OCR outputs being representative of text depicted in a portion of an electronic image. The method may also include identifying a document context associated with the electronic image, and generating an output character set by applying a character resolution model to resolve differences among the plurality of OCR outputs. The character resolution model may define a probability of character recognition accuracy for each of the plurality of OCR engines given the identified document context. The method may also include updating the character resolution model to generate an updated character resolution model such that subsequent generating of output character sets are based on the updated character resolution model.

Claims

1. A method comprising:

receiving, at a processor of a computing system, text outputs from a plurality of optical character recognition (OCR) engines, wherein each of the plurality of OCR engines receives an image of a document and generates an output representative of text depicted in the image of the document;

analyzing, by the processor, the image of the document to identify metadata describing attributes of the document

identifying, by the processor, a difference among the text outputs of the plurality of OCR engines;

resolving, by the processor, the difference among the text outputs of the plurality of OCR engines, by determining a probability of character recognition accuracy for each of the plurality of OCR engines based on the metadata describing the attributes of the document and selecting a character outputted by one of the OCR engines that has a highest probability of character recognition accuracy to be included in an output character set; and

generating, by the processor, the output character set to represent the text in the document.

2. The method of claim 1, wherein determining the probability of character recognition accuracy and selecting the character outputted by the OCR engine that has the highest probability of character recognition accuracy are caused by applying a character resolution model that comprises a Bayesian prior probability distribution.

3. The method of claim 2, further comprising:

updating the character resolution model based on the resolving of the difference to generate an updated character resolution model such that subsequent generating of output character sets is based on the updated character resolution model.

4. The method of claim 1, wherein determining the probability of character recognition accuracy for each of the plurality of OCR engines is based on metadata indicating image resolution of the image of the document.

5. The method of claim 1, wherein the metadata describing the attributes of the document comprises author information associated with the text depicted in the image of the document.

6. The method of claim 1, wherein the metadata describing the attributes of the document comprises language information associated with the text depicted in the image of the document.

7. The method of claim 1, wherein the metadata describing the attributes of the document comprises an image attribute associated with the image of the document, and a content attribute associated with content depicted in the image of the document.

20. The method of claim 1, wherein identifying the difference among the text outputs of the plurality of OCR engines includes:

aligning the text outputs of the plurality of OCR engines with each other on a character by character basis; and

comparing the characters among the text outputs of the plurality of OCR engines to identify differences of the characters among the text outputs of the plurality of OCR engines.

8. A system comprising:

a processor resource; and

a memory storing instructions that when executed by the processor resource cause the processor resource to;

analyze an image of an input document to identify metadata describing attributes of the input document,

receive outputs from a plurality of optical character recognition (OCR) engines, wherein each of the plurality of OCR engines receives the image of the input document and generates an output representative of text depicted in the image of the input document,

identify a difference among the outputs of the plurality of OCR engines,

resolve the difference among the outputs of the plurality of OCR engines based on a character resolution model that utilizes the metadata describing the attributes of the input document and the outputs of the plurality of OCR engines, the character resolution model causing the processor resource to determine a probability of character recognition accuracy for each of the plurality of OCR engines based on the metadata describing the attributes of the input document, and select a character outputted by one of the OCR engines that has a highest probability of character recognition accuracy to be the character for an output document, and

update the character resolution model based on the resolving of the difference to generate an updated character resolution model for subsequent use.

9. The system of claim 8, wherein the character resolution model comprises a Bayesian prior probability distribution.

10. The system of claim 9, wherein the updated character resolution model comprises a Bayesian posterior probability distribution.

11. The system of claim 8, wherein the metadata describing the attributes of the document comprises an image attribute associated with the image of the input document.

12. The system of claim 8, wherein the metadata describing the attributes of the document comprises author information associated with the text depicted in the image of the input document.

13. The system of claim 8, wherein the metadata describing the attributes of the document comprises a content attribute associated with content depicted in the image of the input document.

14. The system of claim 8, wherein the metadata describing the attributes of the document comprises an image attribute associated with the image of the input document, and a content attribute associated with content depicted in the image of the input document.

15. A non-transitory computer-readable storage medium storing instructions that, when executed, cause a processor resource to:

receive outputs from a plurality of optical character recognition (OCR) engines, wherein each of the plurality of OCR engines receives an image of a document and generates an output representative of text depicted in the image of the document;

analyze the image of the document to identify metadata describing attributes of the document

identify a difference among the outputs of the plurality of OCR engines;

resolve the difference among the outputs of the plurality of OCR engines, by determining a probability of character recognition accuracy for each of the plurality of OCR engines based on the metadata describing the attributes of the document and selecting a character outputted by one of the OCR engines that has a highest probability of character recognition accuracy to be included in an output character set; and

generate the output character set to represent the text in the document.

16. The non-transitory computer-readable storage medium of claim 15, wherein the instructions are to cause the processor resource to apply a character resolution model that comprises a Bayesian prior probability distribution to determine the probability of character recognition accuracy and select the character outputted by the OCR engine that has the highest probability of character recognition accuracy.

17. The non-transitory computer-readable storage medium of claim 16, wherein the instructions are to cause the processor resource to update the character resolution model based on the resolving of the difference to generate an updated character resolution model that comprises a Bayesian posterior probability distribution.

18. The non-transitory computer-readable storage medium of claim 15, wherein the metadata describing the attributes of the document comprises an image attribute associated with the image of the document, and a content attribute associated with content depicted in the image of the document.

19. The non-transitory computer-readable storage medium of claim 15, wherein the instructions to cause the processor resource to identify the difference among the outputs of the plurality of OCR engines include instructions to cause the processor resource to:

align the outputs of the plurality of OCR engines with each other on a character by character basis; and

compare the characters among the outputs of the plurality of OCR engines to identify differences of the characters among the outputs of the plurality of OCR engines.