Method and system for fuzzy image classification

• US 5,765,029 A
• Filed: 05/08/1996
• Issued: 06/09/1998
• Est. Priority Date: 05/08/1996
• Status: Expired due to Term

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Schweid, Stuart A., Clark, Raymond J., Shiau, Jeng-Nan
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/646,540
• Time in Patent Office: 762 Days
• Field of Search: 358/462, 382/173, 382/176, 395/61
• US Class Current: 706/52
• CPC Class Codes: G06T 7/44 using image operators, e.g. filters, edge density metrics or local histograms
• CPC Class Codes: H04N 1/40062 Discrimination between different image types, e.g. two-tone, continuous tone
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

A method and system electronically fuzzy classify a pixel belonging to a set of digital image data with respect to a membership of the pixel in a plurality of image classes. This process determines a fuzzy classification of the pixel and generates an effect tag for the pixel based on the fuzzy classification determination. Each class is defined by a set of heuristic rules such that the image classes are non-mutually exclusive. The heuristic rules are a set of conditions that define the membership value of the pixel within a certain class, thereby allowing a pixel to have a membership value in every possible image class.

Claims

1. A method for electronically classify a pixel belonging to a set of digital image data with respect to a membership of the pixel in a plurality of image classes and an other class, comprising the steps of:

(a) receiving a set of digital image data including the pixel;

(b) determining a membership value for the pixel for each image class and a membership value for the pixel for the other class, each image class being defined by a set of heuristic rules, said sets being non-mutually exclusive, the membership value for an image class being independent of the other membership values for the image classes, the membership value for the other class being dependent on the membership values of the image classes; and

(c) generating a classification vector for the pixel based on the determination in said step (b).

2. The method as claimed in claim 1, wherein the set of non-mutually exclusive image classes include text and contone.

3. The method as claimed in claim 1, wherein the set of non-mutually exclusive image classes include text, contone, and halftone.

4. The method as claimed in claim 1, wherein the set of non-mutually exclusive image classes include edge, black, white, pictorial, low frequency halftone, mid frequency halftone, and high frequency halftone.

5. The method as claimed in claim 1, wherein said step (b) includes the substeps of:

(b1) calculating a summation of absolute values of Laplacians of a window of pixels around the pixel;

(b2) determining a grey level range within the window of pixels;

(b3) determining a two-dimensional frequency around the pixel;

(b4) determining a percentage of pixels in the window having a grey value less than a predetermined black threshold; and

(b5) determining a percentage of pixels in the window having a grey value greater than a predetermined white threshold.

6. The method as claimed in claim 5, wherein said step (b) further includes the substeps of:

(b6) determining a conditional value of the pixel with respect to a condition of the summation of said substep (b1) being greater than a first predetermined value, determining a conditional value of the pixel with respect to a condition of the determined range of said substep (b2) being greater than a second predetermined value, and determining a conditional value of the pixel with respect to a condition of the determined frequency of said substep (b3) being greater than a third predetermined value;

(b7) establishing a conditional value for the pixel with respect to a halftone class by determining which conditional value determined in said substep (b6) is a minimum value and assigning the minimum value as the conditional value;

(b8) determining a conditional value of the pixel with respect to a condition of the summation of said substep (b1) being greater than a fourth predetermined value, determining a conditional value of the pixel with respect to a condition of the determined range of said substep (b2) being greater than a fifth predetermined value, and determining a conditional value of the pixel with respect to a condition of the determined frequency of said substep (b3) being greater than the third predetermined value;

edge class by determining which conditional value determined in said substep (b8) is a minimum value and assigning the minimum value as the membership value;

(b10) determining a conditional value of the pixel with respect to a condition of the summation of said substep (b1) being less than the first predetermined value and determining a conditional value of the pixel with respect to a condition of the determined frequency of said substep (b3) being less than the third predetermined value; and

(b11) establishing a conditional value for the pixel with respect to a contone class by determining which conditional value determined in said substep (b10) is a minimum value and assigning the minimum value as the conditional value.

7. The method as claimed in claim 6, wherein said step (b) further includes the substeps of:

(b12) determining a conditional value of the pixel with respect to a condition of the percentage determined in said substep (b4) being greater than a sixth predetermined value;

(b13) establishing a membership value for the pixel with respect to a black class by determining which conditional value determined in said substeps (b11) and (b12) is a minimum value and assigning the minimum value as the membership value;

(b14) determining a conditional value of the pixel with respect to a condition of the percentage determined in said substep (b5) being greater than a seventh predetermined value;

(b15) establishing a membership value for the pixel with respect to a white class by determining which conditional value determined in said substeps (b11) and (b14) is a minimum value and assigning the minimum value as the membership value;

(b16) determining a conditional value of the pixel with respect to a condition of the percentage determined in said substep (b4) being less than the sixth predetermined value and determining a conditional value of the pixel with respect to a condition of the percentage determined in said substep (b5) being less than the seventh predetermined value; and

(b17) establishing a membership value for the pixel with respect to a pictorial class by determining which conditional value determined in said substeps (b11) and (b16) is a minimum value and assigning the minimum value as the membership value.

8. The method as claimed in claim 7, wherein said step (b) further includes the substeps of:

(b18) determining a conditional value of the pixel with respect to a condition of the frequency determined in said substep (b3) being less than an eighth predetermined value;

(b19) establishing a membership value for the pixel with respect to a low frequency halftone class by determining which conditional value determined in said substeps (b7) and (b18) is a minimum value and assigning the minimum value as the membership value;

(b20) determining a conditional value of the pixel with respect to a condition of the frequency determined in said substep (b3) being greater than the eighth predetermined value and determining a conditional value of the pixel with respect to a condition of the frequency determined in said substep (b3) being less than a ninth predetermined value;

(b21) establishing a membership value for the pixel with respect to a mid frequency halftone class by determining which conditional value determined in said substeps (b7) and (b20) is a minimum value and assigning the minimum value as the membership value;

(b22) determining a conditional value of the pixel with respect to a condition of the frequency determined in said substep (b3) being greater than the ninth predetermined value; and

(b23) establishing a membership value for the pixel with respect to a high frequency halftone class by determining which conditional value determined in said substeps (b7) and (b22) is a minimum value and assigning the minimum value as the membership value.

9. The method as claimed in claim 8, wherein said (b) further includes the substeps of:

(b24) establishing a membership value for the pixel with respect to the other class by determining which membership value determined in said substeps (b9), (b13), (b15), (b17), (b19), (b21), and (b23) is a maximum value and assigning a value equal to one minus the maximum value as the membership value for the other.

10. The method as claimed in claim 9, wherein each condition has a predetermined target condition value and two predetermined absolute condition values, the predetermined target condition value and the predetermined absolute condition value are assigned predetermined conditional values, conditional values for condition values between the predetermined target condition value and a predetermined absolute condition value is determined by functionally extrapolating between the assigned predetermined conditional value for the predetermined target condition value and the assigned predetermined conditional value for the predetermined absolute condition value.

11. The method as claimed in claim 10, wherein the functional extrapolation is linear.

12. The method as claimed in claim 10, wherein the assigned predetermined conditional value for the predetermined target condition value is 0.5, and the assigned predetermined conditional values for the predetermined absolute condition values are 0.0 and 1.0.

13. The method as claimed in claim 10, wherein the assigned predetermined conditional value for the predetermined target condition value is 1.0, and the assigned predetermined conditional values for the predetermined absolute condition values are 0.0 and 0.0.

14. The method as claimed in claim 10, wherein the assigned predetermined conditional value for the predetermined target condition value is 0.3, and the assigned predetermined conditional values for the predetermined absolute condition values are 0.0 and 1.0.