FEATURES GENERATION AND SPOTTING METHODS AND SYSTEMS USING SAME

• US 20090060396A1
• Filed: 08/30/2007
• Published: 03/05/2009
• Est. Priority Date: 08/30/2007
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Williamowski, Jutta Katharina, Blessan, Marco
• Granted Patent: US 8,045,798 B2
• Time in Patent Office: Days
• US Class Current: 382/317
• CPC Class Codes: G06V 2201/09 Recognition of logos
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18086 by performing operations within image blocks or by using histograms
• CPC Class Codes: G06V 30/19073 Comparing statistics of pixel or of feature values, e.g. histogram matching

Abstract

An image partitioner is configured to find a partition point that divides a received image into four sub-images each having a pre-selected activated pixel count. A recursion processor is configured to (i) apply the image partitioner to an input image to generate a first partition point and four sub-images and to (ii) recursively apply the image partitioner to at least one of the four sub-images for at least one recursion iteration to generate at least one additional partition point. A formatter is configured to generate a features representation of the input image in a selected format. The features representation is based at least in part on the partition points. The features representation can be used in various ways, such as by a classifier configured to classify the input image based on the features representation.

Claims

1. An image analyzer comprising:

an image partitioner configured to find a partition point that divides a received image into four sub-images each having a pre-selected activated pixel count;

a recursion processor configured to (i) apply the image partitioner to an input image to generate a first partition point and four sub-images and to (ii) recursively apply the image partitioner to at least one of the four sub-images for at least one recursion iteration to generate at least one additional partition point and additional sub-images; and

a formatter configured to generate a features representation of the input image in a selected format, the features representation being based at least in part on the partition points.

2. The image analyzer as set forth in claim 1, wherein the formatter is configured to generate the features representation in a features vector format having at least some dimensions corresponding to coordinates of one or more of the partition points.

3. The image analyzer as set forth in claim 1, wherein the recursion processor is configured to (i) apply the image partitioner to an input image to generate a first partition point and four sub-images and to (ii) recursively apply the image partitioner to all four sub-images for at least one recursion iteration to generate at least four additional partition points.

4. The image analyzer as set forth in claim 3, wherein the recursion processor performs two recursion iterations to generate twenty-one partition points.

5. The image analyzer as set forth in claim 1, wherein the pre-selected activated pixel count is the same for all four sub-images.

6. The image analyzer as set forth in claim 1, wherein the pre-selected activated pixel count is different for at least two of the four sub-images.

7. The image analyzer as set forth in claim 1, wherein the image partitioner comprises:

an integrator configured to compute a two-dimensional integral image from the received image; and

a partition point optimizer configured to iteratively improve an initial partition point to find the partition point that divides the received image into four sub-images each having a pre-selected activated pixel count, the partition point optimizer using pixel values of the two-dimensional integral image to compute a partition point figure-of-merit for each iterative improvement.

8. The image analyzer as set forth in claim 7, wherein the partition point figure-of-merit is operatively equivalent to

9. The image analyzer as set forth in claim 1, further comprising:

a classifier configured to classify the input image based at least on the features representation of the input image.

10. The image analyzer as set forth in claim 9, wherein the classifier is a linear classifier.

11. The image analyzer as set forth in claim 9, wherein the classifier comprises:

an initial fast rejection stage that rejects or provisionally accepts the input image based at least on the features representation of the input image; and

at least one additional stage that rejects or accepts any provisionally accepted input image based on characteristics of the input image.

12. The image analyzer as set forth in claim 11, wherein the at least one additional stage rejects or accepts any provisionally accepted input image based on characteristics of the input image including the features representation of the input image and at least one additional characteristic of the input image.

13. The image analyzer as set forth in claim 9, wherein:

the recursion processor is configured to recursively apply the image partitioner to all four sub-images for at least one recursion iteration to generate at least four additional partition points, and

the formatter is configured to generate the features representation including all of first partition point and the at least four additional partition points in a selected format.

14. The image analyzer as set forth in claim 1, further comprising:

a classifier configured to spot whether the input image corresponds to a selected word based at least on the features representation of the input image.

15. The image analyzer as set forth in claim 1, further comprising:

a classifier configured to spot a whether the input image corresponds to a selected object based at least on the features representation of the input image.

16. The image analyzer as set forth in claim 15, further comprising:

a document image segmentor configured to segment a document image into a plurality of input images that are input to the recursion processor and classifier to determine whether the document image contains the selected object.

17. The image analyzer as set forth in claim 15, wherein the selected object is selected from a group consisting of (i) a selected word, (ii) a selected logo, and (iii) a selected signature.

18. An image analysis method comprising:

finding a partition point that divides an input image into four sub-images each having a pre-selected activated pixel count;

recursively repeating the finding for at least one of the four sub-images one or more times to generate a plurality of partition points; and

classifying the input image based at least in part on the generated partition points.

19. The image analysis method as set forth in claim 18, wherein the recursive repeating is performed at least two times in addition to the initial finding to generate between three and twenty-one partition points inclusive.

20. The image analysis method as set forth in claim 18, wherein the pre-selected activated pixel count is the same for all four sub-images.

21. The image analysis method as set forth in claim 18, wherein the classifying comprises:

performing a fast rejection classification that either rejects or provisionally accepts the input image as corresponding to a selected word, object, or signature; and

conditional upon the fast rejection classification provisionally accepting the input image, performing a second classification of the input image that either rejects or accepts the input image as corresponding to the selected word, object, or signature.

22. The image analysis method as set forth in claim 18, further comprising:

segmenting a document image to define a plurality of input images corresponding to words or objects, the plurality of input images being input to the recursion processor for classification by the classifier.

23. An image analyzer comprising:

an image partitioner configured to find a partition point that divides a received image into four sub-images each having a pre-selected activated pixel count;

a recursion processor configured to recursively apply the image partitioner to an input image to generate a plurality of partition points; and

a classifier configured to classify the input image based at least in part on the partition points generated by the recursion processor.

24. The image analyzer as set forth in claim 23, wherein the classifier is configured to spot whether the input image corresponds to a selected word.

25. The image analyzer as set forth in claim 23, wherein the classifier is configured to spot whether the input image corresponds to a selected signature.

26. The image analyzer as set forth in claim 23, wherein the pre-selected activated pixel count is the same for each of the four sub-images.

27. The image analyzer as set forth in claim 23, wherein the image partitioner comprises:

an integrator configured to compute a two-dimensional integral image from the received image; and

a partition point optimizer configured to iteratively improve an initial partition point based on a figure-of-merit computed using pixel values of the two-dimensional integral image to find the partition point that divides the received image into four sub-images each having a pre-selected activated pixel count.

28. The image analyzer as set forth in claim 23, wherein the recursion processor is configured to recursively apply the image partitioner to the input image for at least two recursive applications to generate P partition points where