Automated detection of pornographic images

• US 20020159630A1
• Filed: 03/29/2001
• Published: 10/31/2002
• Est. Priority Date: 03/29/2001
• Status: Active Grant

• Current Assignee: Cedar Lane Technologies Inc.
• Original Assignee: FotoNation Holdings, LLC (Adeia, Inc.)
• Inventors: Steinberg, Eran, Ciuc, Mihai, Drimbarean, Alexandru, Corcoran, Peter, Grecu, Horia, Buzuloiu, Vasile, Beuran, Razvan
• Granted Patent: US 6,751,348 B2
• Time in Patent Office: Days
• US Class Current: 382/165
• CPC Class Codes: G06T 7/0002 Inspection of images, e.g. flaw detection
• CPC Class Codes: G06T 7/70 Determining position or orientation of objects or cameras camera calibration G06T7/80
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships
• CPC Class Codes: G06V 40/193 Preprocessing; Feature extraction
• CPC Class Codes: H04N 1/624 Red-eye correction control of camera based on recognised facial parts H04N23/611

Abstract

A method of detecting pornographic images, wherein a color reference database is prepared in LAB color space defining a plurality of colors representing relevant portions of a human body. A questionable image is selected, and sampled pixels are compared with the color reference database. Areas having a matching pixel are subjected to a texture analysis to determine if the pixel is an isolated color or if other comparable pixels surround it; a condition indicating possible skin. If an area of possible skin is found, the questionable image is classified as objectionable. A further embodiment includes preparation of a questionable image reference shape database defining objectionable shapes. An image with a detected area of possible skin is compared with the shape database, and depending on the results of the shape analysis, a predefined percentage of the images are classified for manual review.

Claims

1. A method for use in detecting a pornographic image comprising:

analyzing a questionable image, said analyzing including

(a) preparing a final color prototype database having a plurality of final color prototype values, said final color prototype database prepared from selected skin image samples from a plurality of images;

(b) filtering a questionable image, said filtering including

(i) comparing a color of a questionable pixel from a questionable image with said final color prototype values;

(ii) performing a texture analysis on an area surrounding said questionable pixel if said color of said questionable pixel matches a color of a final color prototype, wherein said texture analysis determines a variance in color between said questionable pixel and a color of pixels in said area; and

(iii) classifying said pixel as a potential skin pixel if said texture analysis indicates that said area has a skin texture.

2. A method as recited in claim 1 further comprising

(a) preparing a shape prototype database including a plurality of prototype image shapes;

and

(b) wherein said filtering further includes

(i) comparing a shape of a questionable image including a said potential skin pixel with a said prototype image shape; and

(ii) classifying said questionable image including said potential skin pixel as an objectionable image if a shape of said questionable image corresponds to a shape of a said prototype image shape.

3. A method as recited in claim 1 wherein said final color prototype database includes a plurality of distinct color databases, with each said distinct color database for use in detecting a particular color type.

4. A method as recited in claim 2 wherein said shape prototype database includes a plurality of distinct shape databases wherein each said distinct shape database is for use in detecting a particular shape category.

5. A method as recited in claim 3 wherein a first said distinct color database is configured for detecting at least one of plurality of skin colors.

6. A method as recited in claim 1 further comprising blocking said questionable image in real time from being displayed on a display system if said questionable pixel is determined to be a potential skin pixel.

7. A method as recited in claim 6 wherein said filtering is performed on a server, and said filtering is done prior to a display of said image on said display system.

8. A method as recited in claim 6 wherein said filtering is performed on a user computer.

9. A method as recited in claim 8 wherein said classifying includes preparing classification data indicating if a questionable image is determined to contain skin.

10. A method as recited in claim 9 wherein said classification data is sent from said user computer to a server.

11. A method as recited in claim 4 wherein a first said distinct shape database is for use in detecting a face.

12. A method as recited in claim 4 wherein a first said distinct shape database is for use in detecting a specified body part.

13. A method as recited in claim 12 wherein said body part is genitalia.

14. A method as recited in claim 12 wherein a second said distinct shape database is for use in detecting an erotic position.

15. A method as recited in claim 1 wherein said questionable image is one of a collection of images of digital video.

16. A method as recited in claim 1 wherein said texture analysis includes determining a variance in color in said area surrounding said questionable pixel.

17. A method as recited in claim 1 wherein said texture analysis includes determining a co-occurrence.

18. A method as recited in claim 1 wherein said texture analysis including Gabor filtering.

19. A method as recited in claim 1 wherein said texture analysis including a Fourier Transform based technique.

20. A method as recited in claim 1 wherein each said color prototype is described in a projection of a three dimensional Luminance-Chrominance space into a two dimensional chrominance space.

21. A method as recited in claim 20 wherein each said color prototype is a clusterized group of groups of pixels represented by a center value and a an ellipse defining an area of said clusterized group.

22. A method as recited in claim 21 wherein said clusterized group is prepared using a fuzzy-C-means algorithm.