Word detection method and system

• US 8,224,092 B2
• Filed: 07/08/2008
• Issued: 07/17/2012
• Est. Priority Date: 07/08/2008
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Bressan, Marco J.
• Primary Examiner(s): MARIAM, DANIEL G
• Application Number:
  US12/169,101
• Publication Number:
  US 20100008581A1
• Time in Patent Office: 1,470 Days
• Field of Search: 382/181, 382186-187, 382/190, 382/224, 382/229, 382/177, 382/179, 704/251
• US Class Current: 382/177
• CPC Class Codes: G06V 30/2264 using word shape

Abstract

A method of characterizing a word image includes traversing the word image in steps with a window and at each of a plurality of the steps, identifying a window image. For each of the plurality of window images, a feature is extracted. The word image is characterized, based on the features extracted from the plurality of window images, wherein the features are considered as a loose collection with associated sequential information.

Claims

1. A method of characterizing a word image comprising:

grouping window images of training word images into a set of clusters based on feature similarity;

traversing an acquired word image in steps with a window;

at each of a plurality of the steps, identifying a window image;

for each of the plurality of window images, extracting a feature; and

with a computer processor, characterizing the word image based on the features extracted from the plurality of window images, wherein the features are considered as a loose collection with associated sequential information and the characterizing is based on a similarity between the window image features and those of the clusters.

2. The method of claim 1, wherein the window images are uniformly spaced.

3. The method of claim 1, wherein the window images are irregularly spaced.

4. The method of claim 1, wherein the identifying a window image comprises displacing the window based on a center of gravity of image data within the window, to provide the window image.

5. The method of claim 4, wherein the displacement is on an axis of the traversing direction.

6. The method of claim 4, wherein the displacement is horizontal.

7. The method of claim 4, wherein the window is displaced to the center of gravity.

8. The method of claim 1, wherein the characterization is based on a joint probability of each window image being in a cluster and being at a location in the word image and corresponding joint probabilities of stored training word images.

9. The method of claim 1, wherein the characterizing includes associating location information with the window image features, based on a location of the window image in the word image.

10. The method of claim 9, wherein the associating of location information includes assigning the window image to one of a set of horizontally extending location bins which together span the word image.

11. The method of claim 1, wherein the feature extraction comprises for each window image, determining a features vector;

and the characterizing comprises characterizing the word image based on the features vectors of the plurality of window images.

12. The method of claim 11, further comprising reducing the dimensionality of the features vector.

13. The method of claim 1, further comprising classifying the word image based on the characterization.

14. The method of claim 1, wherein the word image comprises a bitmap acquired by segmenting a document image.

15. The method of claim 1, wherein the word image is characterized without characterization of individual characters of a character string within the word image.

16. A computer program product comprising a non-transitory computer-readable medium encoding instructions, which when executed on a computer causes the computer to perform the method of claim 1.

17. A processing system including a processor which executes instructions stored in memory for performing the method of claim 1.

18. A method of characterizing a word image comprising:

with a computer processor, traversing a word image in steps with a window;

at each of a plurality of the steps, displacing the window based on a center of gravity of image data within the window, to provide a window image;

for each of the plurality of window images, extracting a feature; and

characterizing the word image based on the features extracted from the plurality of window images.

19. The method of claim 18, wherein window images of training words are grouped into a set of clusters based on feature similarity and the characterizing is based on a similarity between the window image features and those of the clusters.

20. A computer program product comprising a non-transitory computer-readable recording medium encoding instructions, which when executed on a computer causes the computer to perform the method of claim 18.

21. A processing system which executes instructions stored in memory for performing the method of claim 18.

22. The method of claim 18, wherein the displacing of the window based on the center of gravity of image data within the window comprises computing a horizontal displacement value from window statistics.

23. The method of claim 18, wherein the displacing of the window based on the center of gravity of image data within the window comprises computing a center of gravity of the image data in the window and moving the window such that it is centered on the center of gravity, the extracting of the feature including computing the feature within the centered window.

24. A processing system comprising:

a document segmentor which processes an input document image to identify word images;

a features extractor which extracts features of an identified word image and computes features vectors therefrom, the features extractor executing instructions for traversing the word image stepwise with a window and, at each of a plurality of the steps, computing a center of gravity of image data in the window, and thereafter displacing the window towards the center of gravity of the image data within the window to provide a window image, whereby window images are irregularly spaced, and, for each of the plurality of window images, computing a feature based on image data within the displaced window; and

a classifier which classifies the word image based on the computed features of the window images and location information for the window images.

25. The processing system of claim 24, wherein the classifier comprises a linear classifier.