System and method for identifying text-based SPAM in rasterized images

• US 7,706,613 B2
• Filed: 08/23/2007
• Issued: 04/27/2010
• Est. Priority Date: 08/23/2007
• Status: Active Grant

• Current Assignee: Kaspersky Lab ZAO
• Original Assignee: Kaspersky Lab ZAO
• Inventors: Smirnov, Evgegy P.
• Primary Examiner(s): PERUNGAVOOR, SATHYANARAYA V
• Application Number:
  US11/844,081
• Publication Number:
  US 20100008579A1
• Time in Patent Office: 978 Days
• Field of Search: 382/182
• US Class Current: 382/182
• CPC Class Codes: G06V 20/95 Pattern authentication; Markers therefor; Forgery detection
• CPC Class Codes: G06V 30/414 Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text

Abstract

A system, method and computer program product for identifying spam in an image, including (a) identifying a plurality of contours in the image, the contours corresponding to probable symbols; (b) ignoring contours that are too small or too large; (c) identifying text lines in the image, based on the remaining contours; (d) parsing the text lines into words; (e) ignoring words that are too short or too long from the identified text lines; (f) ignoring text lines that are too short; (g) verifying that the image contains text by comparing a number of pixels of a symbol color within remaining contours to a total number of pixels of the symbol color in the image, and that there is at least one text line after filtration; and (h) if the image contains text, rendering a spam/no spam verdict based on a contour representation of the text that which appears after step (f).

Claims

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

(a) identifying a plurality of contours in the image, the contours corresponding to probable symbols;

(b) ignoring contours that are too small or too large;

(c) identifying text lines in the image, based on the remaining contours;

(d) parsing the text lines into words;

(e) ignoring words that are too short or too long from the identified text lines;

(f) ignoring text lines that are too short;

(g) verifying that the image contains text by comparing a number of pixels of a symbol color within remaining contours to a total number of pixels of the symbol color in the image; and

(h) if the image contains text, rendering a spam/no spam verdict based on a contour representation of the text that remains after step (f).

2. The method of claim 1, further comprising converting the image from a vector form in a raster form, prior to step (a).

3. The method of claim 1, further comprising converting the image from a multi-color form to a two-color form, prior to step (a).

4. The method of claim 3, further comprising identifying one color of the two colors as a background color, and the other color as the symbol color.

5. The method of claim 4, further comprising generating a spectral distribution of the two-color image, and setting a threshold between the two colors at a level X2+¾

(X1−

X2), X1 being a center of mass of the spectral distribution, and X2 being a first used color in the spectral distribution.

6. The method of claim 4, wherein identification of each contour comprises identifying a first pixel of the symbol color, defining the contour as the first pixel, and then repeatedly expanding the contour in those directions where adjacent pixels are also of the symbol color.

7. The method of claim 1, further comprising, in step (d), for each contour, identifying contours that are within 2Xim of the contour as belonging to a suspected word, wherein Xim is a most frequent distance between all the adjacent contour pairs in all text lines.

8. A system for identifying spam in an image, the system executing the steps of:

(a) identifying a plurality of contours in the image, the contours corresponding to probable symbols;

(b) ignoring contours that are too small or too large;

(c) identifying text lines in the image, based on the remaining contours;

(d) parsing the text lines into words;

(e) ignoring words that are too short or too long, from the identified text lines;

(f) ignoring text lines that are too short;

(g) verifying that the image contains text by comparing a number of pixels of a symbol color within remaining contours to a total number of pixels of the symbol color in the image; and

(h) rendering a spam/no spam verdict based on a contour representation of the remaining text lines and the remaining words.

9. The system of claim 8, wherein the system converts the image from a vector form to a raster form, prior to identifying the contours.

10. The system of claim 8, wherein the system converts the image from a multi-color form to a two-color form, prior to identifying the contours.

11. The system of claim 8, wherein the system identifies one color of the two colors as a background color, and the other color as the symbol color.

12. The system of claim 8, wherein the system generates a spectral distribution of the two-color image, and setting a threshold between the two colors at a level X2+¾

(X1−

X2), X1 being a center of mass of the spectral distribution, and X2 being a first used color in the spectral distribution.

13. The system of claim 12, wherein identification of each contour comprises identifying a first pixel of the symbol color, defining the contour as the first pixel, and then repeatedly expanding the contour in those directions where adjacent pixels are also of the symbol color.

14. The system of claim 8, wherein when rendering each contour, the system identifies contours that are within 2Xim of the contour as belonging to a suspected word, wherein Xim is a most frequent distance between all adjacent contour pairs in all text lines.

15. A computer implemented method for identifying text in an image, the method comprising:

(a) identifying a plurality of contours in the image, the contours corresponding to probable symbols, each contour forming a closed boundary around each probable symbol;

(b) for each contour, identifying adjacent contours that are within 2Xim of the contour, to the left and right, as belonging to the same text line, wherein Xim is a most frequent distance between all adjacent contour pairs in all text lines;

(c) identifying text lines in the image, based on the adjacent contours, wherein adjacent contours belong to the same text line;

(d) ignoring text lines that are too short and parsing the remaining text lines into words; and

(e) identifying presence of text in the image based on the words.

16. The method of claim 15, further comprising converting the image from a multi-color form to a two-color form, prior to step (a).

17. The method of claim 16, further comprising identifying one color of the two colors as a background color, and the other color as the symbol color.

18. The method of claim 17, further comprising generating a spectral distribution of the two-color image, and setting a threshold between the two colors at a level X2+¾

(X1−

X2), X1 being a center of mass of the spectral distribution, and X2 being a first used color in the spectral distribution.

19. The method of claim 18, wherein identification of each contour comprises identifying a first pixel of the symbol color, defining the contour as the first pixel, and then repeatedly expanding the contour in those directions where adjacent pixels are also of the symbol color.