System and method for identifying spam in rasterized images

• US 8,023,697 B1
• Filed: 03/29/2011
• Issued: 09/20/2011
• Est. Priority Date: 03/29/2011
• Status: Expired due to Fees

• Current Assignee: Kaspersky Lab ZAO
• Original Assignee: Kaspersky Lab ZAO
• Inventors: Smirnov, Evgeny P.
• Primary Examiner(s): Johns, Andrew W
• Application Number:
  US13/074,118
• Time in Patent Office: 175 Days
• Field of Search: 382/100, 382/112, 382/176, 382/197, 382/199, 382/286, 709/204, 709/205, 709/206
• US Class Current: 382/100
• CPC Class Codes: G06V 30/418 Document matching, e.g. of document images

Abstract

Disclosed are systems, methods and computer program products for detection of spam in raster images. In one example embodiment, a method comprises identifying objects in the raster image; tracing contours of identified objects; computing, angles of inclination of tangents at plurality of points on an object contour; determining, based on the computed angles of inclination, one or more attributes of the object contour, including lengths of line segments of the contour, angles between the line segments, lengths of arcs the contour and radii of curvature of the arcs; generating object signatures containing one or more attributes of the object contours; comparing object signatures of the image with known spam image signatures; and determining that the image contains spam if the number of object signatures that are substantially similar to the known spam image signatures exceeds a predetermined threshold.

Claims

1. A computer implemented method for identifying spam in an image, comprising:

identifying one or more objects in the image, including text and graphic objects;

tracing contours of one or more identified objects;

computing angles of inclination of tangents at plurality of points on an object contour;

determining, based on the angles of inclination, one or more attributes of the object contour, including lengths of line segments of the contour, angles between line segments the contour, lengths of arcs of the contour and radii of curvature of the arcs of the contour;

generating object signatures containing one or more attributes of the object contours;

comparing signatures of one or more objects of the image with known spam signatures;

determining that the image contains spam if the number of object signatures that are substantially similar to the known spam signatures exceeds a predetermined threshold.

2. The method of claim 1, further comprising converting the image from a vector format into a raster format, prior identifying a plurality of objects in the image.

3. The method of claim 1, further comprising converting the image from a multi-color format into a grayscale format, prior to identifying a plurality of objects in the image.

4. The method of claim 1, wherein identifying one or more objects in the image further comprises rejecting one or more small objects in the image.

5. The method of claim 1, wherein tracing an image contour comprises:

identifying a background as a plurality of pixels having a first shade of gray;

identifying one or more pixels that have a second shade of gray;

defining the contour as the first pixel having the second shade of gray; and

repeatedly expanding the contour of the object in those directions where pixels adjacent to the background pixels are of the second shade of gray.

6. The method of claim 1, wherein computing angles of inclination of tangents at plurality of points on an object contour comprises:

selecting a point on the object contour;

constructing a tangent at the selected point; and

computing angle of inclination of the tangent with respect to horizontal.

7. The method of claim 1, wherein determining, based on the angles of inclination, one or more attributes of the object contour comprises:

determining deviation in the values of the angles of inclination of tangents at the two or more adjacent points on the object contour; and

if the deviation is substantially zero, the two or more adjacent points form a straight line segment, and the length of the line segment is calculated;

if the deviation is substantially constant, the two or more adjacent points form an arc, and the length and radius of curvature of the arc is calculated.

8. A system for identifying spam in an image, comprising:

a memory being configured to store an image; and

a processor coupled to the memory, the processor being configured to;

identify one or more objects in the image, including text and graphic objects;

trace contours of one or more identified objects;

compute angles of inclination of tangents at plurality of points on an object contour;

determine, based on the angles of inclination, one or more attributes of the object contour, including lengths of line segments of the contour, angles between line segments the contour, lengths of arcs of the contour and radii of curvature of the arcs of the contour;

generate object signatures containing one or more attributes of the object contours;

compare signatures of one or more objects of the image with known spam signatures; and

determine that the image contains spam if the number of object signatures that are substantially similar to the known spam signatures exceeds a predetermined threshold.

9. The system of claim 8, wherein the processor being further configured to convert the image from a vector format into a raster format, prior to identifying a plurality of objects in the image.

10. The system of claim 8, wherein the processor being further configured to convert the image from a multi-color format into a grayscale format, prior to identifying a plurality of objects in the image.

11. The system of claim 8, wherein to identify one or more objects in the image, the processor being further configured to reject one or more small objects in the image.

12. The system of claim 8, wherein to trace an image contour, the processor being further configured to:

identify a background as a plurality of pixels having a first shade of gray;

identify one or more pixels that have a second shade of gray;

define the contour as the first pixel having the second shade of gray; and

repeatedly expand the contour of the object in those directions where pixels adjacent to the background pixels are of the second shade of gray.

13. The system of claim 8, wherein to compute angles of inclination of tangents at plurality of points on an object contour, the processor being further configured to:

select a point on the object contour;

construct a tangent at the selected point; and

compute angle of inclination of the tangent with respect to horizontal.

14. The system of claim 8, wherein to determine, based on the angles of inclination, one or more attributes of the object contour, the processor being further configured to:

determine deviation in the values of the angles of inclination of tangents at the two or more adjacent points on the object contour; and

if the deviation is substantially zero, the two or more adjacent points form a straight line segment, and the length of the line segment is calculated;

if the deviation is substantially constant, the two or more adjacent points form an arc, and the length and radius of curvature of the arc is calculated.

15. A computer program product embedded in a non-transitory computer-readable storage medium, the computer-readable storage medium comprising computer-executable instructions for identifying spam in an image, the medium includes instructions for:

identifying one or more objects in the image, including text and graphic objects;

tracing contours of one or more identified objects;

computing angles of inclination of tangents at plurality of points on an object contour;

determining, based on the angles of inclination, one or more attributes of the object contour, including lengths of line segments of the contour, angles between line segments the contour, lengths of arcs of the contour and radii of curvature of the arcs of the contour;

generating object signatures containing one or more attributes of the object contours;

comparing signatures of one or more objects of the image with known spam signatures;

determining that the image contains spam if the number of object signatures that are substantially similar to the known spam signatures exceeds a predetermined threshold.

16. The medium of claim 15, further comprising instructions for converting the image from a vector format into a raster format, prior to identifying a plurality of objects in the image.

17. The medium of claim 15, further comprising instructions for converting the image from a multi-color format into a grayscale format, prior to identifying a plurality of objects in the image.

18. The medium of claim 15, wherein instructions for tracing an image contour further comprise instructions for:

identifying a background as a plurality of pixels having a first shade of gray;

identifying one or more pixels that have a second shade of gray;

defining the contour as the first pixel having the second shade of gray; and

repeatedly, expanding the contour of the object in those directions where pixels adjacent to the background pixels are of the second shade of gray.

19. The medium of claim 15, wherein instructions for computing angles of inclination of tangents at plurality of points on an object contour further comprise instructions for:

selecting a point on the object contour;

constructing a tangent at the selected point; and

computing angle of inclination of the tangent with respect to horizontal.

20. The medium of claim 15, wherein instructions for determining, based on the angles of inclination, one or more attributes of the object contour further comprise instructions for:

determining deviation in the values of the angles of inclination of tangents at the two or more adjacent points on, the object contour; and

if the deviation is substantially zero, the two or more adjacent points form a straight;

line segment, and the length of the line segment is calculated;

if the deviation is substantially constant, the two or more adjacent points form an arc, and the length and radius of curvature of the arc is calculated.