Fuzzy image segmentation

• US 6,549,656 B1
• Filed: 04/02/1996
• Issued: 04/15/2003
• Est. Priority Date: 11/29/1993
• Status: Expired due to Fees

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Lin, Ying-wei
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US08/626,433
• Time in Patent Office: 2,569 Days
• Field of Search: 382/173, 382/176, 382/155, 382/159, 382/224, 382/228, 345/113, 345/115, 358/462, 358/467, 395/61, 395/1, 706/1, 706/900, 706/902
• US Class Current: 382/173
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

A digital image processing system for automatically segmenting a set of input image signals into a combination of predetermined classes of imagery, the set of input image signals forming part of a video image generated by an image input terminal. The system includes a classification circuit that receives the set of input image signals and classifies them as a ratio of the predetermined classes of imagery. A plurality of image processing circuits also receives the set of input image signals, each of which is adapted to process a unique class of imagery selected from the predetermined classes of imagery to generate a set of output image signals for that predetermined class of imagery. Finally, a mixing circuit combines each of the sets of output image signals determined by the plurality of image processing circuits in accordance with the ratio determined by the classification circuit to form a single set of output image signals, the mixing circuit minimizing the generation of artifacts formed in the single set of output image signals by providing a gradual transition between each of the predetermined class of imagery.

Claims

1. A digital image processing system for automatically classifying a set of input image signals into a combination of at least three predetermined classes of imagery and producing output image signals in accordance with the classes, the set of input image signals forming part of a video image generated by an image input terminal, comprising:

a mixing circuit;

a data buffer for receiving the set of input image signals;

a classification circuit for characterizing properties of each of the input signals, said classification circuit producing, as a function of the properties of each of the input image signals, a plurality of non-zero probability values, each non-zero probability value representing a likelihood that one of the input image signals of the set of input image signals is a member of a respective one of the at least three predetermined classes of imagery, and transmitting probability signals indicative thereof to said mixing circuit; and

a plurality of image processing circuits receiving the set of input image signals from the data buffer, each of said plurality of image processing circuits being adapted to process the input image signals in accordance with a process identified for one of the at least three predetermined classes of imagery, said mixing circuit combining the signals from at least two of said plurality of image processing circuits in accordance with the probability signals received from said classification circuit to form a single set of output image signals, wherein a subset of the output image signals, representing input image signals from a transition zone where there is no high degree of certainty associated with any of the predetermined classes of imagery, comprise signals processed by the at least two of said plurality of image processing circuits so as to account for gradual shifts between regions of the input image representing the different classes of imagery.

2. The system as described claim 1, wherein said classification circuit comprises:

a memory;

a characteristic equation stored in said memory, said characteristic equation being derived from a population of video images that represent an image property having a relationship with the predetermined classes of imagery;

a characteristic calculator adapted to calculate a characteristic value for the set of input image signals of the image property represented by said characteristic equation; and

a probability classifier for evaluating said characteristic equation using the characteristic value calculated by said characteristic calculator to determine a probability value for each of the predetermined classes of imagery.

3. The system as described in claim 2, wherein said classification circuit comprises an input control for adjusting said characteristic equation.

4. The system as described in claim 2, wherein the image property described by said characteristic equation is determined from a peak count of the input image signals.

5. The system as described in claim 2, wherein said characteristic equation defines a unique probability distribution for each predetermined class of imagery.

6. The system described in claim 2, wherein said mixing circuit comprises:

a plurality of multiplier circuits; and

an adder circuit, coupled to said plurality of multiplier circuits, for providing a single set of unquantized output image signals.

7. The system described in claim 6, wherein said mixing circuit comprises a quantization circuit, coupled to said adder circuit, for quantizing said single set of unquantized output image signals.

8. The system as described in claim 1, wherein the predetermined classes of imagery include low frequency halftones, high frequency halftones and continuous tones.

9. The system as described in claim 1, further comprising an image output terminal for receiving the single set of output image signals.

10. The system as described in claim 9, wherein said image output terminal is a printer.

11. The system as described in claim 9, wherein the image input terminal is a scanner.

12. The system as described in claim 1, further comprising a memory for receiving and storing the input image signals.

13. A method for automatically classifying a set of input image signals into combinations of at least three predetermined image classes, the set of input image signals forming part of a video image generated by an image input terminal, comprising:

classifying the input image signals by image type, including identifying properties of each input image signal within the set;

determining from the identified properties for each input image signal the likelihood of each input image signal being a member of one of the at least three predetermined image classes;

producing a plurality of non-zero probability values for each input image signal, each non-zero probability value representing the likelihood that the input image signal is a member of a respective one of the at least three predetermined image classes;

operating on the set of input image signals with a plurality of image processing circuits, each of the image processing circuits being uniquely adapted for processing the input image signals in accordance with a process associated with a single predetermined image class, to provide a plurality of sets of output image signals wherein each set corresponds to classes; and

combining output image signals from the plurality of sets of output image signals, in accordance with the non-zero probability values associated with each of the input image signals, to form a single set of output image signals, wherein input image signals in a transition zone found between different classes of imagery within the video image are combined to effect a shift in processing methods so as to reduce the appearance of any artifacts at corresponding transition zones in the output image.

14. The method according to claim 13, wherein said classifying step comprises the steps of:

deriving a characteristic equation from a population of video images, the characteristic equation relating an image property of the population of video images with the predetermined image classes;

calculating a characteristic value for the set of input image signals, the characteristic value representing the image property represented by the characteristic equation; and

evaluating the characteristic equation using the characteristic value to determine a probability value for each of the predetermined image classes.

15. The method according to claim 14, wherein said classifying step comprises the step of adjusting the characteristic equation.

16. The method according to claim 14, wherein said classifying step comprises the step of providing the characteristic equation with a unique probability distribution for each predetermined image class.

17. The method according to claim 14, wherein said classifying step classifies the image property described by the characteristic equation determined by the peak count of the input image signals.

18. The method according to claim 13, wherein the processing step processes the predetermined classes of imagery in accordance with image processing circuits adapted for processing low frequency halftones, high frequency halftones and continuous tones.

19. The method according to claim 13, wherein said combining step comprises the steps of:

multiplying each sets of output image signals with a fraction defined by the probability of each of the predetermined classes of imagery; and

adding the products of said multiplying step to provide an unquantized single set of output image signals.

20. The method according to claim 19, wherein said combining step comprises the step of quantizing the unquantized single set of output image signals.

21. The method according to claim 13, further comprising the step of transmitting the single set of output image signals to an image output terminal for reproduction.

22. The method according to claim 13, wherein said transmitting step transmits the single set of output image signals to a printer.

23. The method according to claim 13, further comprising the step of scanning a document with an image scanner to generate input image signals.

24. The method according to claim 13, further comprising the step of storing the set of input image signals provided by the image input terminal in a memory.