Adversarial approach for identifying inappropriate text content in images

• US 8,098,939 B2
• Filed: 05/16/2007
• Issued: 01/17/2012
• Est. Priority Date: 12/04/2006
• Status: Active Grant

• Current Assignee: Trend Micro Inc.
• Original Assignee: Trend Micro Inc.
• Inventors: Oliver, Jonathan James
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US11/803,963
• Publication Number:
  US 20080131005A1
• Time in Patent Office: 1,707 Days
• Field of Search: 382/229, 382/203, 382/217, 382/218
• US Class Current: 382/229
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/414 Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text
• CPC Class Codes: H04L 51/212 using filtering or selective blocking

Abstract

An adversarial approach in detecting inappropriate text content in images. An expression from a listing of expressions may be selected. The listing of expressions may include words, phrases, or other textual content indicative of a particular type of message. Using the selected expression as a reference, the image is searched for a section that could be similar to the selected expression. The similarity between the selected expression and the section of the image may be in terms of shape. The section may be scored against the selected expression to determine how well the selected expression matches the section. The score may be used to determine whether or not the selected expression is present in the image.

Claims

1. A method of identifying inappropriate text content in images, the method to be performed using a computer and comprising:

selecting an expression from a listing of expressions;

extracting an image from a message;

using the selected expression as a reference, finding a section of the image that corresponds to a start point and an end point of the selected expression;

comparing the section of the image to the selected expression to determine how well the section of the image matches the selected expression; and

determining if the selected expression is present in the section of the image based on the comparison of the image to the selected expression;

wherein comparing the section of the image to the selected expression comprises assigning, to pixel blocks between the start point and the end point of the section of the image, probabilities of being characters of the selected expression, and further comprising deeming the message as spam based on presence of the selected expression in the section of the image.

2. The method of claim 1 wherein finding the section of the image that corresponds to the start point and the end point of the selected expression comprises:

finding a section of the image that has a shape similar to that of the selected expression.

3. The method of claim 1 wherein comparing the section of the image against the selected expression comprises scoring the image against the selected expression to generate a score.

4. The method of claim 3 wherein determining if the selected expression is present in the section of the image based on the comparison of the image to the selected expression comprises comparing the score to a threshold.

5. The method of claim 4 wherein the selected expression is deemed present in the section of the image when the score is higher than the threshold.

6. The method of claim 1 further comprising performing OCR on the image and wherein the section of the image comprises a text output of the OCR on the image.

7. The method of claim 1 further comprising building a text string representing a text content of the section without performing OCR on the section.

8. The method of claim 7 wherein comparing the section of the image to the selected expression comprises comparing the text string to the selected expression.

9. The method of claim 1 wherein the message comprises an email and the selected expression comprises a word or phrase indicative of spam.