Contextualizing noisy samples by substantially minimizing noise induced variance

• US 8,948,535 B2
• Filed: 09/02/2010
• Issued: 02/03/2015
• Est. Priority Date: 01/27/2010
• Status: Active Grant

• Current Assignee: DST Technologies, Inc. (SS&C Technologies Holdings Incorporated)
• Original Assignee: DST Technologies, Inc. (SS&C Technologies Holdings Incorporated)
• Inventors: Ives, Paul M., Hsu, Chin
• Primary Examiner(s): TUCKER, WESLEY J
• Application Number:
  US12/874,573
• Publication Number:
  US 20110182527A1
• Time in Patent Office: 1,615 Days
• Field of Search: None
• US Class Current: 382/264
• CPC Class Codes: G06V 30/40 Document-oriented image-based pattern recognition
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables
• CPC Class Codes: G06V 30/418 Document matching, e.g. of document images

Abstract

A system for contextualizing noisy samples by substantially minimizing noise induced variance may include a memory, an interface, and a processor. The memory is operative to store exemplars. The processor is operative to receive, via the interface, a sample which includes exemplar content corresponding to one of the exemplars, and noise. Variance induced by the noise may differentiate the sample from one or more of the exemplars. The processor may generalize the sample and the exemplars in order to substantially minimize the variance. The processor may compare the generalized sample to the generalized exemplars to identify the exemplar corresponding to the exemplar content of the sample. The processor may contextualize the sample based on a document type of the identified exemplar. The processor may present the contextualized sample to a user to facilitate interpretation thereof, and in response thereto, receive data representative of a user determination associated with the noise.

Claims

1. A method of contextualizing noisy samples comprising:

receiving a sample comprising exemplar content corresponding to one of a plurality of exemplars and noise, wherein the noise comprises a modification to the exemplar content relative to the exemplar content as originally comprised by a physical object from which the sample was generated, and wherein variance induced by the noise differentiates the sample from one or more of the plurality of exemplars;

generalizing the sample and the plurality of exemplars in order to minimize the variance induced by the noise;

comparing the generalized sample to each of the plurality of generalized exemplars to identify which of the plurality of exemplars corresponds to the exemplar content of the sample;

contextualizing the sample based on a document type corresponding to the identified exemplar of the plurality of exemplars; and

presenting the contextualized sample to a user to facilitate interpretation thereof and, in response thereto, receiving data representative of a user determination associated with the noise.

2. The method of claim 1 wherein the sample comprises an electronic document image.

3. The method of claim 2 wherein generalizing the sample and the plurality of exemplars further comprises applying a moving average filter to the sample and the plurality of exemplars in order to minimize the variance induced by the noise.

4. The method of claim 2 wherein the noise comprises handwritten information, scanner skew, image noise, or a combination thereof.

5. The method of claim 2 further comprising:

determining multiple exemplars of the plurality of exemplars which may correspond to the sample;

presenting the sample and the multiple exemplars of the plurality of exemplars to a user to facilitate determination thereof and, in response thereto, receiving data representative of a user determination identifying which of the multiple exemplars of the plurality of exemplars corresponds to the sample;

contextualizing the sample based on a document type corresponding to the identified exemplar; and

presenting the contextualized sample to a user to facilitate interpretation thereof and, in response thereto, receiving data representative of a user determination associated with the noise.

6. The method of claim 1 further comprising providing a confidence indication which indicates a level of confidence that the sample corresponds to the one of the plurality of exemplars, wherein the level of confidence is based on a size of the noise relative to the size of the sample.

7. The method of claim 1 wherein the sample comprises at least one of video or audio.

8. The method of claim 1 wherein at least one of the plurality of exemplars comprises a contextualized sample or sample characterized by a known document type.

25. The method of claim 1 wherein the noise comprises handwritten information.

9. A method of contextualizing noisy samples comprising:

receiving a sample electronic document image comprising exemplar information corresponding to one of a plurality of exemplar electronic document images and random information, wherein the random information comprises a modification to the exemplar information relative to the exemplar information as originally comprised by a physical object from which the sample electronic document image was generated, and wherein variance induced by the random information renders the sample electronic document image distinguishable from one or more of the plurality of exemplar electronic document images;

applying a filter to the sample electronic document image and the plurality of exemplar electronic document images in order to minimize the variance induced by the random information of the sample electronic document image;

identifying the one of the plurality of exemplar electronic document images corresponding to the exemplar information of the sample electronic document image by comparing the filtered sample electronic document image to each of the plurality of filtered exemplar electronic document images;

contextualizing the sample electronic document image based on a document type corresponding to the identified exemplar electronic document image; and

presenting the contextualized sample electronic document image to a user to facilitate interpretation thereof and, in response thereto, receiving data representative of a user determination associated with the random information.

10. A method of contextualizing noisy samples comprising:

receiving a sample electronic document image comprising exemplar content and noise, wherein the noise comprises a modification to the exemplar content relative to the exemplar content as originally comprised by a physical object from which the sample electronic document image was generated;

generating a plurality of transformations of the sample electronic document image, wherein each of the plurality of transformations of the sample electronic document image corresponds to a different level of filtering of the exemplar content and the noise;

comparing the plurality of transformations of the sample electronic document image to a plurality of transformations of each of a plurality of exemplar electronic document images;

determining one of the plurality of exemplar electronic document images of which a greatest number of the plurality of transformations satisfy a matching criteria with respect to the plurality of transformations of the sample electronic document image having a same level of filtering;

contextualizing the sample electronic document image based on a document type corresponding to the determined one of the plurality of exemplar electronic document images; and

presenting the contextualized sample electronic document image to a user to facilitate interpretation thereof and, in response thereto, receiving data representative of a user determination associated with the noise.

11. The method of claim 10 wherein comparing the plurality of transformations of the sample electronic document image to the plurality of transformations of each of the plurality of exemplar electronic document images further comprises:

(a) identifying a transformation of the plurality of transformations of the sample electronic document image having a highest level of filtering which has not been compared;

(b) identifying transformations for each of the plurality of exemplar electronic document images corresponding to the highest level of filtering identified in (a);

(c) comparing the transformations for each of the plurality of exemplar electronic document images identified in (b) to the transformation of the sample electronic document image identified in (a);

(d) eliminating any of the plurality of exemplar electronic document images from the comparing when the transformations identified in (b) for any of the plurality of exemplar electronic document images do not satisfy a matching criteria with respect to the transformation of the sample electronic document image identified in (a); and

(e) repeating steps (a)-(d) until only one of the plurality of exemplar electronic document images remains.

12. The method of claim 10 wherein generating the plurality of transformations of the sample electronic document image, wherein each of the plurality of transformations of the sample electronic document image corresponds to the different level of filtering of the exemplar content and the noise further comprises generating the plurality of transformations of the sample electronic document image by recursively applying a moving average filter to the sample electronic document image.

13. The method of claim 10 wherein each of the plurality of transformations of the sample electronic document image comprises a different resolution.

14. The method of claim 10 wherein each of the plurality of transformations of the sample electronic document image comprises a different level of compression relative to the sample electronic document image.

15. A method of classifying an electronic document image, the method comprising:

receiving a sample electronic document image comprising content and noise, wherein the noise comprises a modification to the content relative to the content as originally comprised by a physical object from which the sample electronic document image was generated;

filtering the sample electronic document image to account for variance induced by the noise;

filtering a the plurality of exemplar electronic document images to a same level of filtering as the sample electronic document image;

determining one of the plurality of exemplar electronic document images corresponding to the content of the sample electronic document image by comparing the filtered sample electronic document image to each of the plurality of filtered exemplar electronic document images;

contextualizing the sample electronic document based on a document type corresponding to the determined one of the plurality of exemplar electronic document images; and

presenting the contextualized sample electronic document image to a user to facilitate interpretation thereof and, in response thereto, receiving data representative of a user determination associated with the noise.

16. A method of classifying an electronic document image, the method comprising:

identifying a plurality of exemplar electronic document images wherein each of the plurality of exemplar electronic document images is characterized by a document type;

receiving a sample electronic document image comprising a variation of one of the plurality of exemplar electronic document images, wherein the variation comprises a modification to content as originally comprised by a physical object from which the sample electronic document image was generated, and wherein variance induced by the variation distinguishes the sample electronic document image from one or more of the plurality of exemplar electronic document images;

filtering the sample electronic document image and the plurality of exemplar electronic document images in order to minimize the variance induced by the variation;

comparing the filtered sample electronic document image to each of the plurality of filtered exemplar electronic document images to determine one of the plurality of exemplar electronic document images corresponding to the sample electronic document image;

contextualizing the sample electronic document based on the document type corresponding to the determined one of the plurality of exemplar electronic document images; and

presenting the contextualized sample electronic document image to a user to facilitate interpretation thereof and, in response thereto, receiving data representative of a user determination associated with the variation.

17. A system for contextualizing noisy samples comprising:

a memory operative to store a plurality of exemplars;

an interface coupled with the memory and operative to receive a sample comprising exemplar content corresponding to one of the plurality of exemplars and noise, wherein the noise comprises a modification to the exemplar content relative to the exemplar content as originally comprised by a physical object from which the sample was generated; and

a processor coupled with the interface and operative to receive, via the interface, the sample comprising the exemplar content corresponding to one of the plurality of exemplars and the noise, wherein variance induced by the noise differentiates the sample from one or more of the plurality of exemplars, generalize the sample and the plurality of exemplars to minimize the variance induced by the noise, compare the generalized sample to each of the plurality of generalized exemplars to identify which of the plurality of exemplars corresponds to the exemplar content of the sample, contextualize the sample based on a document type corresponding to the identified exemplar, and present the contextualized sample to a user to facilitate interpretation thereof, and in response thereto, receive data representative of a user determination associated with the noise.

18. The system of claim 17 wherein the sample comprises an electronic document image.

19. The system of claim 18 wherein the processor is further operative to apply a filter to the sample and the plurality of exemplars in order to minimize the variance induced by the noise.

20. The method of claim 19 wherein the filter applied by the processor comprises a moving average filter.

21. The system of claim 17 wherein the noise comprises handwritten information, scanner skew, image noise, or a combination thereof.

22. The system of claim 17 wherein the processor is further operative to determine multiple exemplars of the plurality of exemplars which may correspond to the exemplar content of the sample, present the sample and the multiple exemplars of the plurality of exemplars to a user to facilitate determination thereof and, in response thereto, receive data representative of a user determination of which of the multiple of the plurality of exemplars corresponds to the exemplar content of the sample, contextualize the sample based on a document type of the identified exemplar, and present the contextualized sample to a user to facilitate interpretation thereof and, in response thereto, receive data representative of a user determination associated with the noise.

23. The system of claim 17 wherein the processor is further operative to provide a confidence indication which indicates a level of confidence that the sample corresponds to the one of the plurality of exemplars, wherein the level of confidence is based on a size of the noise relative to the size of the sample.

24. The system of claim 17 wherein the sample comprises at least one of video or audio.

26. The system of claim 17 wherein the noise comprises handwritten information.