Color printer characterization using optimization theory and neural networks

• US 6,480,299 B1
• Filed: 11/25/1998
• Issued: 11/12/2002
• Est. Priority Date: 11/25/1997
• Status: Expired due to Term

• Current Assignee: The Board of Regents of the University of Colorado (University of Colorado System)
• Original Assignee: University Technology Corporation
• Inventors: Subbarayan, Ganesh, Drakopoulos, Paul Apostolos
• Primary Examiner(s): WALLERSON, MARK E
• Application Number:
  US09/199,626
• Time in Patent Office: 1,448 Days
• Field of Search: 358/1.1, 358/1.5, 358/1.6, 358/1.12, 358/1.13, 358/1.14, 358/1.15, 358/1.16, 358/1.9
• US Class Current: 358/1.9
• CPC Class Codes: G06F 3/12 Digital output to print unit , e.g. line printer, chain printer
• CPC Class Codes: H04N 1/6033 using test pattern analysis H04N1/6055 takes precedence
• CPC Class Codes: H04N 1/6058 Reduction of colour to a range of reproducible colours, e.g. to ink- reproducible colour gamut

Abstract

A color management method/apparatus generates image color matching and International Color Consortium (ICC) color printer profiles using a reduced number of color patch measurements. Color printer characterization, and the generation of ICC profiles usually require a large number of measured data points or color patches and complex interpolation techniques. This invention provides an optimization method/apparatus for performing LAB to CMYK color space conversion, gamut mapping, and gray component replacement. A gamut trained network architecture performs LAB to CMYK color space conversion to generate a color profile lookup table for a color printer, or alternatively, to directly control the color printer in accordance with the a plurality of color patches that accurately. represent the gamut of the color printer. More specifically, a feed forward neural network is trained using an ANSI/IT-8 basic data set consisting of 182 data points or color patches, or using a lesser number of data points such as 150 or 101 data points when redundant data points within linear regions of the 182 data point set are removed. A 5-to-7 neuron neural network architecture is preferred to perform the LAB to CMYK color space conversion as the profile lookup table is built, or as the printer is directly controlled. For each CMYK signal, an ink optimization criteria is applied, to thereby control ink parameters such as the total quantity of ink in each CMYK ink printed pixel, and/or to control the total quantity of black ink in each CMYK ink printed pixel.

Claims

1. A method of controlling a color printer having a gamut that defines all colors that can be printed by the printer, the printer being connected in a computer system that contains a set of printer independent input signals that command the printer to print an output print that corresponds to the set of input signals, the method comprising the steps of:

using the printer to print a plurality of output prints said plurality of output prints being representative of the gamut of the printer and comprising in the range of from about 101 to about 182 output prints;

providing a neural network comprising in the range of from about 5 to about 7 neurons;

using said plurality of output prints to train said neural network and thereby provide a gamut trained neural network for converting printer dependent color signals into printer independent color signals that correspond to the gamut of the printer;

using said gamut trained neural network to generate a printer profile lookup table tabulating a series of printer independent color signals to a series of printer dependent color signals that correspond to the gamut of the printer;

using said set of printer independent input signals to address said printer profile lookup table; and

controlling the printer in accordance with a set of printer dependent color signals that correspond to the set of printer independent input signals.

2. The method of claim 1 wherein said neurons operate to implement a non-linear hyperbolic tangent activation function.

3. The method of claim 1 wherein said device-independent color signals comprise LAB signals and wherein said device-dependent color signals comprise CMYK signals.

4. The method of claim 1 wherein said plurality of output prints representative of the gamut of the printer are selected in accordance with an ANSII basic data set in xy chromaticity coordinates.

5. The method of claim 4 wherein said device-independent color signals comprise LAB signals and wherein said device-dependent color signals comprise CMYK signals.

6. A method for controlling a color printer having a color gamut, said color printer being responsive to a computer system that provides color space image signals that command said color printer to print an output print that corresponds to said color space image signals, the method comprising the steps of:

producing a plurality of output prints from said color printer representative of the color gamut of said color printer and comprising a plurality of different colors selected in accordance with an ANSI print set in xy chromaticity coordinates;

analyzing said plurality of output prints;

providing a neural network;

training said neural network using the results of said analysis step to thereby provide a gamut trained neural network;

using said gamut trained neural network to produce a printer profile lookup table that correlates color space signals to the color gamut of said color printer;

connecting said color space image signals to address said printer profile lookup table; and

controlling said color printer in accordance with a portion of the color gamut of said color printer that corresponds to said color space image signals.

7. The method of claim 6 wherein said plurality of output prints comprise in the range of from about 101 to about 182 output prints, and wherein said neural network comprises in the range of from about 5 to about 7 neurons.

8. The method of claim 7 wherein said neurons operate to implement a non-linear hyperbolic tangent activation function.

9. The method of claim 6 wherein said color space image signals are printer independent, and wherein the color gamut of said color printer defines printer dependent signals.

10. The method of claim 6 wherein said neural network operates to implement a non-linear hyperbolic tangent activation function.

11. A method of making a profile lookup table for a CMYK color printer comprising the steps of:

producing a relatively small number of different color prints using the CMYK printer;

using said color prints to train a neural network and thereby define a gamut of the CMYK printer as said gamut is defined by said relatively small number of different color prints, said neural network comprising from about 5 to about 7 neurons; and

using said trained neural network to generate a printer profile lookup table that converts LAB values to corresponding CMYK values in accordance with said gamut of the CMYK printer.

12. The method of claim 11 wherein said step of using said color prints to train said neural network includes the step of using an optimization routine to constrain said corresponding CMYK values in accordance with preselected criteria relationship between the C, M, Y and K colors.

13. The method of claim 11 further including, the steps of:

providing LAB input signals that define a color image to be printed by the CMYK printer;

using said LAB signals as an input to said lookup table; and

providing CMYK signals from said lookup table to the CMYK printer corresponding to said LAB signals, such that output prints are produced bv the CMYK printer as defined by said RGB input signals.

14. Apparatus for producing a profile lookup table for use by a CMYK color printer in producing color prints, the profile lookup table comprising a plurality of table entries wherein each entry comprises a LAB value and a corresponding CMYK value in accordance with a print gamut of the CMYK color printer, the apparatus comprising:

a neural network defining the print gamut of the CMYK color printer;

said neural network having an input for receiving a test CMYK value;

said neural network having an output for generating a LAB value that corresponds to said print gamut and to said test CMYK value;

an error determining network;

said error determining network having a first input connected to receive a series of LAB values wherein each of said received LAB values corresponds to a table entry of the profile lookup table;

said error determining network having a second input connected to said output of said neural network to receive test LAB values;

said error determining network having an error output that is responsive to a difference in the LAB values received by said first and second inputs;

an error testing network establishing a LAB error criteria as a function of C, M, Y and K variables;

said error testing network having an input connected to said error output of said error determining network;

said error testing network having a NO output indicative of failure of said error output of said error determining network to meet said LAB error criteria;

said error testing network having a YES output indicative of said error output of said error determining network meeting said LAB error criteria; and

said YES output providing a CMYK value table entry to the profile lookup table that corresponds to a LAB value then connected to said first input of said error determining network.

15. The apparatus of claim 14 including:

an optimization network connected intermediate to said NO output of said error testing network and said input of said neural network;

said optimization network operating to minimize said error output of said error determining network that is responsive to a difference in the LAB values that are received by said first and second inputs of said error determining network.

16. The apparatus claim 15 wherein said neural network includes from 5 to 7 neurons.

17. The apparatus of claim 16 wherein said optimization network operates to apply a preselected constraint on said CMYK value table entries to the profile lookup table.

18. The apparatus of claim 17 wherein said preselected constraint is that the total quantity of printing ink that is represented by each CMYK value table entry shall not exceed a preselected quantity.

19. Apparatus for producing a profile lookup table for use by a color printer in producing color prints, the profile lookup table comprising a plurality of table entries wherein each entry comprises a printer independent color value and a corresponding printer dependent color value in accordance with a print gamut of the color printer, the apparatus comprising:

a neural network defining the print gamut of the color printer;

said neural network having an input for receiving a test printer dependent color value;

said neural network having an output for generating a printer independent color value that corresponds to said print gamut and to said test printer dependent color value;

an error determining network;

said error determining network having a first input connected to receive a series of printer independent color values wherein each of said received printer independent color values corresponds to a table entry of the profile lookup table;

said error determining network having a second input connected to said output of said neural network to receive test printer independent color values therefrom;

said error determining network having an error output that is responsive to a difference in printer independent color values received by said first and second inputs;

an error testing network establishing a color value error criteria;

said error testing network having an input connected to said error output of said error determining network;

said error testing network having a NO output indicative of failure of said error output of said error determining network to meet said color value error criteria;

said error testing network having a YES output indicative of said error output of said error determining network meeting said color value error criteria; and

said YES output providing a printer dependent color value table entry to the profile lookup table that corresponds to a printer independent color value then connected to said first input of said error determining network.

20. The apparatus of claim 19 including:

an optimization network connected intermediate said NO output of said error testing network and said input of said neural network;

said optimization network operating to minimize said error output of said error determining network that is responsive to a difference in the printer independent color values that are received by said first and second inputs of said error determining network.

21. The apparatus claim 20 wherein said neural network includes from 5 to 7 neurons.

22. The apparatus of claim 21 wherein said optimization network operates to apply a preselected ink constraint on said printer dependent color value table entries.

23. The apparatus of claim 22 wherein said preselected constraint is that the total quantity of printing ink/specified by each printer dependent color value table entry shall not exceed a preselected quantity of ink.

24. The apparatus of claim 19 wherein said printer dependent color value table entrvy is a CYMK table entry, and wherein said preselected constraint relates to the quantity of K within each CMYK table entry.

25. Apparatus for converting printer independent color signals to printer dependent color signals, comprising:

an input for receiving a printer independent color signal that defines a color pixel image to be printed by a color printer;

a color gamut conversion network having an input for receiving a test printer dependent color signal, and having an output for providing a test printer independent color signal in accordance with a color gamut of the color printer and in accordance with said received test printer dependent color signal;

an error determining junction having a first input connected to said input receiving said printer independent color signal defining a color pixel image to be printed by the color printer, having a second input connected to said output of said color gamut conversion network for receiving said test printer independent color signal, and having an output responsive to a difference between said printer independent color signal defining a color pixel image to be printed by the color printer and said test printer independent color signal;

a decision network constructed and arranged to implement a function f=C1(error signal)+C2(a parameter of a printed color pixel), wherein said terms “

C1” and

“

C2”

are constants, wherein said term “

error signal”

is a function of an acceptable color error of a printed color pixel corresponding to said printer independent color signal, and wherein said term “

a parameter of a printed color pixel”

relates to a factor, such as the ink cost of a printed color pixel corresponding to said input printer independent signal;

said decision network having an input connected to said output of said error determining junction, having a YES output responding to said function being satisfied, and having a NO output responding to said function not being satisfied;

a generating network for generating a test printer dependent color signal;

said generating network having an input connected to said NO output of said decision network, and said generating network operating to generate said test printer dependent signal on an output of said generating network in response to said function not being satisfied; and

circuit means connecting said output of said generating network to said input of said color gamut conversion network.

26. The apparatus of claim 25 wherein said color gamut conversion network comprises a neural network that is trained in accordance with said color gamut of the color printer.

27. The apparatus of claim 25 including:

means responsive to said YES output of said decision network operating to provide said test printer dependent color signal to control the printer as said function of said input printer independent color signal.

28. The apparatus of claim 27 wherein said color gamut conversion network comprises a neural network that is trained in accordance with said color gamut of the color printer.

29. The apparatus of claim 25 wherein so long as said NO output of said decision network continues to persist, said generating network operates to generate a string of individually different test printer dependent signals until such time as a test printer dependent signal is generated that results in said function being satisfied.

30. The apparatus of claim 29 wherein a serial string of said input printer independent color signals are applied to said first input of said error determining junction, and including means responsive to a corresponding string of YES outputs from said decision network operable to build a printer profile lookup table having a string of entries, each individual entry being in accordance with an individual input printer independent color signal and a corresponding individual test printer dependent signals that resulted in said function being satisfied.

31. The apparatus of claim 30 wherein said color gamut conversion network comprises a neural network that is trained in accordance with said color gamut of the color printer.

32. The apparatus of claim 25 wherein said printer independent color signals are LAB signals, wherein said printer dependent color signals are CMYK signals, and wherein the color printer is a CMYK printer.

33. The apparatus of claim 32 wherein said YES output of said decision network is operable to apply a test CMYK signal that resulted in said YES output being produced to control said CMYK printer.

34. The apparatus of claim 33 wherein said color gamut conversion network comprises a neural network that is trained in accordance with said color gamut of the CMYK printer.

35. The apparatus of claim 32 wherein said YES output of said decision network is operable to apply an entry to a profile lookup table of said CMYK printer, said entry comprising a test CMYK signal that resulted in said YES output and said input LAB signal.

36. The apparatus of claim 35 wherein said color gamut conversion network comprises a neural network that is trained in accordance with said color gamut of the CMYK printer.

37. A method for converting an input color signal to an output color signal that correspond to the color gamut of a color printer, comprising the steps of:

receiving said input color signal;

providing color gamut conversion means for generating a test color signal for comparison to said input color signal, said test color signal being in accordance with said color gamut of said color printer;

providing error determining means for determining an error magnitude of a difference between said input color signal and said test color signal;

providing decision means responsive to said error magnitude, said decision means operating to determine an acceptable error or an unacceptable error;

providing first means responsive to said unacceptable error for controlling said color gamut conversion means to generate a different test color signal for comparison to said input color signal;

repeating operation of said first means until a given test color signal results in an acceptable error being determined by said decision means; and

providing second means responsive to said acceptable error for generating said output color signal in accordance with said given test color signal.

38. The method of claim 37 wherein said output color signal is operable to produce a printed pixel.

39. The method of claim 37 wherein said output color signal is applied as an entry to a color printer profile lookup table.

40. The method of claim 37 wherein said output color signal defines a color pixel to be printed by a color printer, and wherein said decision means implements the function;

acceptable error=C1(actual error)+C2(cost of said color pixel),

wherein said terms “

C1” and

“

C2”

are constants, wherein said term “

actual error”

is a function of an acceptable visual color error of said color pixel, and wherein said term “

cost of said color pixel”

relates to the ink cost of said color pixel.

41. The method of claim 40 wherein said color gamut conversion means comprises a neural network that is trained in accordance with a plurality of color prints that define said color gamut.

42. The method of claim 40 wherein said input color signal is a LAB signal, wherein said test color signal is a LAB signal, wherein said output color signal is a CMYK signal, and wherein said color printer is a CMYK printer.

43. The method of claim 42 wherein said color gamut conversion means comprises a neural network that is trained in accordance with a plurality of CMYK color prints that define said color gamut.

44. The method of claim 37 including the steps of:

generating a string of input color signals and then, by operation of the method of claim 37, generating a corresponding string of output color signals for which acceptable errors have been determined; and

generating a profile lookup table for said color printer in accordance with said string of input color signals and said corresponding string of output color signals.

45. The method of claim 44 wherein said input color signals are LAB signals, wherein said test color signals are LAB signals, wherein said output color signals are CMYK signals, wherein said profile lookup table comprises a string of table entries, each table entry of which relates an individual LAB signal to a corresponding individual CMYK signal, and wherein said color printer is a CMYK printer.

46. The method of claim 45 wherein said color gamut conversion means comprises a neural network that is trained in accordance with a plurality of CMYK color prints that define said color gamut.

47. Apparatus for constructing a profile lookup table for use by a given CMYK printer in printing a CMYK color print in response to a plurality of printer independent color signals that define the CMYK color print, the profile lookup table containing a plurality of individual entries, each entry containing a printer independent signal and a corresponding CMYK signal in accordance with a gamut of the given CMYK printer, the profile lookup table for use in converting printer independent signals to CMYK signals in accordance with the gamut of the given CMYK printer, said apparatus comprising:

a neural network having an input and an output;

said neural network being responsive to a CMYK signal at said input and operating to provide a printer independent signal at said output in accordance with the gamut of the given CMYK printer;

apparatus input means for receiving a plurality of input printer independent signals, each individual one of said input printer independent signals corresponding to an individual entry in the profile lookup table;

error measuring means having a first input connected to receive an individual one of said plurality of input printer independent signals, having a second input connected to said output of said neural network, and having a difference error output that is responsive to a difference between signals applied to said first and second inputs;

error criteria means having an input connected to said difference error output of said error measuring means, said error criteria means operating to set CMYK optimization criteria to be met by all CMYK entries to the profile lookup table;

said error criteria means providing a NO output so long as said difference between signals applied to said first and second inputs of said error measuring means does not satisfy said CMYK optimization criteria;

said error criteria means providing a YES output when said difference between signals applied to said first and second inputs of said error measuring means satisfies said CMYK optimization criteria;

first means responsive to said NO output operating to sequentially apply a different CMYK signal to said input of said neural network until such time as a given CMYK signal applied to said input of said neural network operates to provide said YES output from said error criteria means; and

second means responsive to said YES output for producing an entry in the profile lookup table that comprises said individual one of said input printer independent signal and said given CMYK signal.

48. The apparatus of claim 47 wherein said CMYK optimization criteria includes a total ink quantity and/or a total black ink quantity represented by said given CMYK signal.

49. The apparatus of claim 48 wherein said printer independent signals are LAB signals.

50. Apparatus for controlling a given CMYK printer to print a CMYK color print in response to a plurality of printer independent color signals that define a plurality of color pixels within the CMYK color print, said CMYK printer having a CMYK color gamut, said apparatus comprising:

gamut-defining means having an input and an output;

said gamut-defining means being responsive to a CMYK signal at said input and operating to provide a printer independent signal at said output in accordance with the CMYK gamut of the given CMYK printer;

apparatus input means for receiving the plurality of input printer independent signals that define a plurality of pixels within the CMYK color print, each individual one of said input printer independent signals corresponding to an individual pixel of the pixels within the CMYK color print;

error measuring means having a first input connected to receive each individual one of said plurality of input printer independent signals, having a second input connected to said output of said gamut-defining means, and having a difference error output that is responsive to a difference between signals applied to said first and second inputs;

error criteria means having an input connected to said difference error output of said error measuring means, said error criteria means operating to set CMYK optimization criteria to be met by all color pixels of the CMYK color print;

said error criteria means providing a NO output so long as said difference between signals applied to said first and second inputs of said error measuring means does not satisfy said CMYK optimization criteria;

said error criteria means providing a YES output when said difference between signals applied to said first and second inputs of said error measuring means satisfies said CMYK optimization criteria;

first means responsive to said NO output operating to sequentially apply a different CMYK signal to said input of said gamut-defining means, until such time as a given CMYK signal applied to said input of said gamut-defining means operates to provide said YES output from said error criteria means; and

second means responsive to said YES output for providing said given CMYK signal to control the given printer.

51. The apparatus of claim 50 wherein said CMYK optimization criteria includes a total ink quantity and/or a total black ink quantity represented by said given CMYK signal.

52. The apparatus of claim 51 wherein said printer independent signals are LAB signals.

53. The apparatus of claim 52 wherein said gamut defining means is a neural network.