Methods of Content-Based Image Identification

• US 20140003712A1
• Filed: 12/20/2012
• Published: 01/02/2014
• Est. Priority Date: 06/28/2012
• Status: Active Grant

• Current Assignee: Lexmark International Incorporated (Ninestar Corporation)
• Original Assignee: Lexmark International Incorporated (Ninestar Corporation)
• Inventors: Eid, Ahmed Hamad Mohamed, Cholewo, Tomasz Jan
• Granted Patent: US 9,064,316 B2
• Time in Patent Office: Days
• US Class Current: 382/164
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1478 of characters or characters lines
• CPC Class Codes: G06V 30/18095 Summing image-intensity values; Projection and histogram analysis

Abstract

A method of identifying one or more image features of an image based on content of the image according to one example embodiment includes receiving an image, correcting a distortion and a tone of the image, segmenting the corrected image and extracting the one or more image features, generating an input image descriptor based on the one or more extracted image features, matching the input image descriptor with a reference image descriptor, and performing an action based on the matched reference image descriptor.

Claims

1. A method of identifying one or more image features of an image based on content of the image, comprising:

receiving an image;

correcting a distortion and a tone of the image;

segmenting the corrected image and extracting the one or more image features;

generating an input image descriptor based on the one or more extracted image features;

matching the input image descriptor with a reference image descriptor; and

performing an action based on the matched reference image descriptor.

2. The method of claim 1, wherein the received image is captured by a mobile device.

3. The method of claim 1, wherein correcting the tone of the image includes performing local analysis to determine a document background of the image and contrast values in the image.

4. The method of claim 3, wherein correcting the tone of the image further includes applying varying tone correction in different areas of the image.

5. The method of claim 1, wherein correcting the tone of the image includes remapping areas in the image to a same constant value.

6. The method of claim 1, wherein segmenting the corrected image includes segmenting the image into one of a foreground and a background class.

7. The method of claim 6, further comprising analyzing the foreground class to define at least one connected component.

8. The method of claim 7, further comprising classifying the at least one connected component into one of a text and a non-text component.

9. The method of claim 7, further comprising classifying the at least one connected component into one of an edge and a non-edge image content.

10. The method of claim 7, wherein the input image descriptor is a histogram of percentage area of each connected component relative to areas corresponding to a bounding box of the connected component.

11. The method of claim 1, wherein the input image descriptor is a histogram of a count of gradient magnitudes at different angles.

12. The method of claim 1, wherein the input image descriptor is identified textual information from the image.

13. The method of claim 12, wherein the identified textual information is an output of optical character recognition of the image.

14. The method of claim 1, wherein the action performed is retrieving stored information corresponding to the input image descriptor.

15. A method of identifying one or more image features of an image captured by a mobile device based on content of the image, comprising:

extracting the one or more image features from the image;

generating an input image descriptor based on the one or more extracted image features;

selecting a matching reference image descriptor for the input image descriptor from a plurality of reference image descriptors, the matching reference image descriptor having a highest matching score with the input image descriptor of the plurality of reference image descriptors; and

performing a predetermined action based on the matching reference image descriptor.

16. The method of claim 15, wherein the action performed includes retrieving information associated with the extracted image features.

17. The method of claim 15, wherein the action performed includes routing the image to a designated path.

18. A method of identifying one or more image features of an image captured by a mobile device based on content of the image, comprising:

extracting the one or more image features from the image;

generating an input image descriptor based on the one or more extracted image features;

selecting a reference image descriptor from a set of reference image descriptors that matches the input image descriptor; and

performing an action based on the matching reference image descriptor,

wherein the matching reference image descriptor is selected based on a determined correlation coefficient of the input image descriptor and the matching reference image descriptor from the set of reference image descriptors.

19. The method of claim 18, wherein the action performed includes retrieving information associated with the extracted image features.

20. The method of claim 18, wherein the action performed includes producing a printed output of the image.