Evaluating graphic image files for objectionable content

• US 20050232480A1
• Filed: 05/03/2005
• Published: 10/20/2005
• Est. Priority Date: 05/26/2000
• Status: Active Grant

• Current Assignee: Dana B. Swift
• Original Assignee: Dana B. Swift
• Inventors: Swift, Dana B.
• Granted Patent: US 7,027,645 B2
• Time in Patent Office: Days
• US Class Current: 382/165
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 16/5862 using texture
• CPC Class Codes: G06V 10/56 relating to colour
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

Computer based method and apparatus for evaluating a graphic image file for the presence of objectionable visual content. The graphic image file comprises a number of pixels each having spectral components defined in accordance with a color model (such as RGB). Relative weighting of the spectral components of each pixel is performed to classify each pixel as either not representing human skin or as possibly representing human skin. For those sets of adjacent pixels that possibly represent human skin, a brightness test is performed to determine whether the set of pixels represents a curved surface. A statistical analysis evaluates the run length of each such set of pixels. Display of the graphic image file is allowed or disallowed based on the results of the foregoing analyses. Web pages available from the World Wide Web are further screened for objectionable terms in the associated URLs and text.

Claims

1. A computer based method for evaluating a graphic image file for the presence of objectionable visual content, the graphic image file comprising a number of pixels each having spectral components defined in accordance with a color model, the method comprising steps of:

(a) evaluating relative weighting of the spectral components of each pixel in turn to classify each said pixel as either not representing human skin or as possibly representing human skin; and

(b) processing the graphic image file in relation to the resulting classification of each said pixel during the evaluating step (a);

wherein the color model comprises a numeric scheme so that the spectral components of each pixel are expressed as the equivalent of a set of red, green and blue (R, G, B) components corresponding to the respective amounts of red, green and blue in each pixel; and

further wherein each of the (R, G, B equivalent) components has a numeric value ranging from m to n, and wherein the evaluating step (a) comprises a step of comparing the respective numeric values of the (R, G, B equivalent) components; and

further wherein said pixel is classified as not representing human skin when at least one of the following conditions is satisfied;

the numeric value of the R,G,B equivalent blue component is greater than the numeric value of the R,G,B equivalent red component;

the numeric value of the R,G,B equivalent blue component is greater than the numeric value of the R,G,B equivalent green component;

the numeric value of the R,G,B equivalent green component is greater than the numeric value of the R,G,B equivalent red component;

the numeric value of the R,G,B equivalent blue component is less than one-quarter the numeric value of the R,G,B equivalent red component; and

the numeric value of the R,G,B equivalent blue component exceeds a threshold T.

2. A computer based method for filtering a web page for objectionable content, the web page available from the World Wide Web and having a URL address, text words and at least one graphic image file comprising a number of pixels each having spectral components defined in accordance with a color model, the method comprising steps of:

(a) providing a list of marker terms;

(b) comparing a requested URL for a selected web page with the list of marker terms and disallowing display of the selected web page when at least one term in the requested URL appears in the list of marker terms;

(c) comparing the text words from the selected web page to the list of marker terms and disallowing display of the selected web page when at least one of the text words appears in the list of marker terms;

(d) evaluating a graphic image file in the selected web page for the presence of objectionable visual content by evaluating relative weighting of the spectral components of each pixel in turn to classify each said pixel as either not representing human skin or as possibly representing human skin;

(e) allowing display of the selected web page when each said pixel is classified as not representing human skin;

(f) counting a number v of graphic image files in the selected web page;

(g) performing the evaluating step (d) on each of the v of graphic image files in turn;

(h) identifying a subset w of the v graphic image files having at least one pixel classified as possibly representing human skin; and

(i) displaying the selected web page when w is less than a selected percentage of v.

3. A computer based method for filtering a web page for objectionable content, the web page available from the World Wide Web and having a URL address, text words and at least one graphic image file comprising a number of pixels each having spectral components defined in accordance with a color model, the method comprising steps of:

(a) providing a list of marker terms;

(b) comparing a requested URL for a selected web page with the list of marker terms and disallowing display of the selected web page when at least one term in the requested URL appears in the list of marker terms;

(c) comparing the text words from the selected web page to the list of marker terms and disallowing display of the selected web page when at least one of the text words appears in the list of marker terms;

(d) evaluating a graphic image file in the selected web page for the presence of objectionable visual content by evaluating relative weighting of the spectral components of each pixel in turn to classify each said pixel as either not representing human skin or as possibly representing human skin; and

(e) allowing display of the selected web page when each said pixel is classified as not representing human skin;

wherein the color model comprises an RGB color model so that the spectral components of each pixel are expressed as a set of red, green and blue (R, G, B) components corresponding to the respective amounts of red, green and blue in each pixel; and

further wherein each of the (R,G,B) components has a numeric value ranging from m to n; and

further wherein the evaluating step (d) comprises a step of comparing the respective numeric values of the (R,G,B) components; and

further wherein said pixel is classified as not representing human skin when at least one of the following conditions is satisfied;

the numeric value of the blue component is greater than the numeric value of the red component;

the numeric value of the blue component is greater than the numeric value of the green component;

the numeric value of the green component is greater than the numeric value of the red component;

the numeric value of the blue component is less than one-quarter the numeric value of the red component; and

the numeric value of the blue component exceeds a threshold T.

4. The method of claim 3, further comprising a step of:

(f) disallowing display of the selected web page when at least one pixel is classified as possibly representing human skin.

5. The method of claim 3, further comprising steps of counting the number of text words from the selected web page, and allowing display of the selected page irrespective of the presence of at least one graphic image file when no term in the requested URL appears in the list of marker terms, none of the text words appears in the list of marker terms and the number of text words exceeds a predetermined number.

6. The method of claim 3, further comprising steps of:

(f) counting a number v of graphic image files in the selected web page;

(g) performing the evaluating step (d) on each of the v of graphic image files in turn;

(h) identifying a subset w of the v graphic image files having at least one pixel classified as possibly representing human skin; and

(i) displaying the selected web page when w is less than a selected percentage of v.

7. A data recording medium which stores a software program configured to carry out the method of claim 3 when loaded into computer memory.

8. A computer based method for filtering a web page for objectionable content, the web page available from the World Wide Web and having a URL address, text words and at least one graphic image file comprising a number of pixels each having spectral components defined in accordance with a color model, the method comprising steps of:

(a) providing a list of marker terms;

(b) comparing a requested URL for a selected web page with the list of marker terms and disallowing display of the selected web page when at least one term in the requested URL appears in the list of marker terms;

(c) comparing the text words from the selected web page to the list of marker terms and disallowing display of the selected web page when at least one of the text words appears in the list of marker terms;

(d) evaluating a graphic image file in the selected web page for the presence of objectionable visual content by evaluating relative weighting of the spectral components of each pixel in turn to classify each said pixel as either not representing human skin or as possibly representing human skin; and

(e) allowing display of the selected web page when each said pixel is classified as not representing human skin;

wherein the evaluating step (d) further comprises steps of performing a brightness analysis of each set of adjacent pixels classified as possibly representing human skin to determine whether each said set of adjacent pixels represents a curved surface, and allowing display of the graphic image file when each said set of adjacent pixels is determined to not represent a curved surface;

and further wherein the brightness analysis of performing step (c) comprises steps of arranging each said set of adjacent pixels into a leading portion, a middle portion and a trailing portion of pixels, and comparing the spectral components of the pixels in the middle portion to the spectral components of the pixels in the leading and trailing portions.

9. A computer based method for filtering a web page for objectionable content, the web page available from the World Wide Web and having a URL address, text words and at least one graphic image file comprising a number of pixels each having spectral components defined in accordance with a color model, the method comprising steps of:

(a) providing a list of marker terms;

(b) comparing a requested URL for a selected web page with the list of marker terms and disallowing display of the selected web page when at least one term in the requested URL appears in the list of marker terms;

(c) comparing the text words from the selected web page to the list of marker terms and disallowing display of the selected web page when at least one of the text words appears in the list of marker terms;

(d) evaluating a graphic image file in the selected web page for the presence of objectionable visual content by evaluating relative weighting of the spectral components of each pixel in turn to classify each said pixel as either not representing human skin or as possibly representing human skin; and

(e) allowing display of the selected web page when each said pixel is classified as not representing human skin;

wherein the evaluating step (d) further comprises steps of performing a brightness analysis of each set of adjacent pixels classified as possibly representing human skin to determine whether each said set of adjacent pixels represents a curved surface, and allowing display of the graphic image file when each said set of adjacent pixels is determined to not represent a curved surface;

wherein the evaluating step (d) further comprising steps of performing a run length analysis by determining the number of pixels in each set of adjacent pixels classified as possibly representing human skin, and disallowing display of the graphic image file when at least one said set of adjacent pixels has a number of pixels that is greater than a selected percentage of the maximum number of possible adjacent pixels in a row of the graphic image file.

10. A computer system configured to evaluate a graphic image file for the presence of objectionable visual content, the graphic image file comprising a number of pixels each having spectral components defined in accordance with a color model, the computer system comprising:

a general purpose computer having associated programming to;

(a) evaluate relative weighting of the spectral components of each pixel in turn to classify each said pixel as either not representing human skin or as possibly representing human skin;

(b) allow display of the graphic image file when all pixels are classified as not representing human skin; and

(c) disallow display of the graphic image file when at least some pixels are classified as possibly representing human skin;

wherein the color model comprises an RGB color model so that the spectral components of each pixel are expressed as a set of red, green and blue (R, G, B) components corresponding to the respective amounts of red, green and blue in each pixel; and

further wherein said pixel is classified as not representing human skin when at least one of the following conditions is satisfied;

the numeric value of the blue component is greater than the numeric value of the red component;

the numeric value of the blue component is greater than the numeric value of the green component;

the numeric value of the green component is greater than the numeric value of the red component;

the numeric value of the green component is greater than the numeric value of the red component;

the numeric value of the blue component is less than one-quarter the numeric value of the red component; and

the numeric value of the blue component exceeds a threshold T;

wherein the computer system is connected to a computer network and configured to receive web pages from the World Wide Web each having a URL address and each potentially having text words and at least one graphic image file, wherein the general purpose computer is further programmed to;

(d) compare a requested URL for a selected web page with a list of marker terms and disallow display of the selected web page when at least one term in the requested URL appears in the list of marker terms; and

(e) compare the text words from the selected web page to the list of marker terms and disallow display of the selected web page when at least one of the text words appears in the list of marker terms.

11. A computer system configured to evaluate a graphic image file for the presence of objectionable visual content, the graphic image file comprising a number of pixels each having spectral components defined in accordance with a color model, the computer system comprising:

a general purpose computer having associated programming to;

(a) evaluate relative weighting of the spectral components of each pixel in turn to classify each said pixel as either not representing human skin or as possibly representing human skin;

(b) allow display of the graphic image file when all pixels are classified as not representing human skin; and

(c) disallow display of the graphic image file when at least some pixels

are classified as possibly representing human skin;

wherein the associated programming further configures the general purpose computer to perform a brightness analysis of each set of adjacent pixels classified as possibly representing human skin to determine whether each said set of adjacent pixels represents a curved surface, and allow display of the graphic image file when each said set of adjacent pixels is determined to not represent a curved surface; and

further wherein the associated programming further configures the general purpose computer to perform a run length analysis by determining the number of pixels in each set of adjacent pixels classified as possibly representing human skin, and disallowing display of the graphic image file when at least one said set of adjacent pixels has a number of pixels that is greater than a selected percentage of the maximum number of possible adjacent pixels in a row of the graphic image file;

and further wherein the computer system is connected to a computer network and configured to receive web pages from the World Wide Web each having a URL address and each potentially having text words and at least one graphic image file, wherein the general purpose computer is further programmed to;

(d) compare a requested URL for a selected web page with a list of marker terms and disallow display of the selected web page when at least one term in the requested URL appears in the list of marker terms; and

(e) compare the text words from the selected web page to the list of marker terms and disallow display of the selected web page when at least one of the text words appears in the list of marker terms.