Method and apparatus for estimating a special reflectance distribution or a spectral transmittance distribution using illuminant-independent characteristic parameters

• US 5,864,834 A
• Filed: 03/23/1995
• Issued: 01/26/1999
• Est. Priority Date: 05/20/1994
• Status: Expired due to Fees

• Current Assignee: TOYO Ink Manufacturing Company Limited
• Original Assignee: TOYO Ink Manufacturing Company Limited
• Inventors: Arai, Yoshihumi
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Shah, Sanjiv
• Application Number:
  US08/409,016
• Time in Patent Office: 1,405 Days
• Field of Search: 395/22, 395/38, 395/80, 395/131, 706/16, 706/12, 356/402, 345/431
• US Class Current: 706/16
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology

Abstract

A neural network is made to undergo learning such that at least three characteristic parameters, which correspond to an inputted set of color information values when the set of color information values, including at least three color information values, is inputted and which are obtained by multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution, are outputted. A subject set of color information values is transformed into the characteristic parameters by using the neural network which has completed learning, and a spectral reflectance distribution or a spectral transmittance distribution is estimated by linear polynomial approximation using the transformed characteristic parameters, eigenvectors obtained by the multivariate analysis, and a mean vector of the spectral reflectance distribution or the spectral transmittance distribution.

Claims

1. A method of estimating a spectral reflectance distribution or a spectral transmittance distribution of a color with a set of device values used for color reproduction with a color reproduction apparatus, including at least three device values used for color reproduction with a color reproduction apparatus, on the basis of the set of device values used for color reproduction with a color reproduction apparatus, comprising the steps of:

causing a neural network to undergo learning such that at least three principal components, which correspond to the inputted set of device values used for color reproduction with a color reproduction apparatus when the set of device values used for color reproduction with a color reproduction apparatus is inputted and which are obtained by multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution, are outputted;

transforming a subject set of device values used for color reproduction with a color reproduction apparatus into the principal components by using said neural network which has completed learning; and

estimating a spectral reflectance distribution or a spectral transmittance distribution by linear polynomial approximation using the transformed characteristics parameters, eigenvectors obtained by the multivariate analysis, and a mean vector of the spectral reflectance distribution or the spectral transmittance distribution.

2. A method of estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 1, wherein the learning step is effected by:

measuring the spectral reflectance distribution of each a multiplicity of color chips corresponding to the at least three device values used for color reproduction with a color reproduction apparatus and each having a color whose density differs for each of the device values used for color reproduction with a color reproduction apparatus;

determining an eigenvector by performing principal-component analysis by using a covariance matrix of values in each of which the mean vector of the spectral reflectance distribution is subtracted from the spectral reflectance distribution;

determining the at least three principal components corresponding to each of sets of device values used for color reproduction with a color reproduction apparatus, including at least three device values used for color reproduction with a color reproduction apparatus, by using the determined eigenvector and the values in each of which the mean vector of the spectral reflectance distribution is subtracted from the spectral reflectance distribution; and

causing said neural network to undergo learning by using the sets of device values used for color reproduction with a color reproduction apparatus and the principal components such that the at least three characteristics parameters corresponding to the set of device values used for color reproduction with a color reproduction apparatus are outputted when the set of device values used for color reproduction with a color reproduction apparatus including the at least three device values used for color reproduction with a color reproduction apparatus is inputted.

3. A method of estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 1, wherein the set of device values used for color reproduction with a color reproduction apparatus is RGB values which are three primary colors of additive color mixture, CMY values which are three primary colors of subtractive primary colors, or CMYK values.

4. A method of estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 1, wherein said neural network is a multilayered feedforward neural network comprising three to five layers.

5. A method of estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 1, wherein KL expansion is used as the linear polynomial approximation, and coefficients of KL expansion are used as the principal components.

6. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution of a color with a set of device values used for color reproduction with a color reproduction apparatus, including at least three device values used for color reproduction with a color reproduction apparatus, on the basis of the set of device values used for color reproduction with a color reproduction apparatus, comprising:

a neural network for learning such that at least three principal components, which correspond to the inputted set of device values used for color reproduction with a color reproduction apparatus when the set of device values used for color reproduction with a color reproduction apparatus is inputted and which are obtained by multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution, are outputted; and

estimating means for estimating a spectral reflectance distribution or a spectral transmittance distribution by linear polynomial approximation using the transformed characteristics parameters, eigenvectors obtained by the multivariate analysis, and a mean vector of the spectral reflectance distribution or the spectral transmittance distribution.

7. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 6, wherein said neural network

measures the spectral reflectance distribution of each a multiplicity of color chips corresponding to the at least three device values used for color reproduction with a color reproduction apparatus and each having a color whose density differs for each of the device values used for color reproduction with a color reproduction apparatus;

determines an eigenvector by performing principal-component analysis by using a covariance matrix of values in each of which the mean vector of the spectral reflectance distribution is subtracted from the spectral reflectance distribution;

determines the at least three principal components corresponding to each of sets of device values used for color reproduction with a color reproduction apparatus, including at least three device values used for color reproduction with a color reproduction apparatus, by using the determined eigenvector and the values in each of which the mean vector of the spectral reflectance distribution is subtracted from the spectral reflectance distribution; and

undergoes learning by using the sets of device values used for color reproduction with a color reproduction apparatus and the principal components such that the at least three characteristics parameters corresponding to the set of device values used for color reproduction with a color reproduction apparatus are outputted when the set of device values used for color reproduction with a color reproduction apparatus, including the at least three device values used for color reproduction with a color reproduction apparatus, is inputted.

8. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 6, wherein the set of device values used for color reproduction with a color reproduction apparatus is RGB values which are three primary colors of additive color mixture, CMY values which are three primary colors of subtractive primary colors, or CMYK values.

9. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 6, wherein said neural network is a multilayered feedforward neural network comprising three to five layers.

10. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 6, wherein KL expansion is used as the linear polynomial approximation, and coefficients of KL expansion are used as the principal components.

11. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution of a color with a set of device values used for color reproduction with a color reproduction apparatus, including at least three device values used for color reproduction with a color reproduction apparatus on the basis of the set of device values used for color reproduction with a color reproduction apparatus, comprising:

input means to which a set of device values used for color reproduction with a color reproduction apparatus, including at least three device values used for color reproduction with a color reproduction apparatus, is inputted and which converts the inputted set of device values used for color reproduction with a color reproduction apparatus into corresponding electrical signals;

a storage device for storing eigenvectors obtained by multivariate analysis, a mean vector of the spectral reflectance distribution or the spectral transmittance distribution, coefficients of connection learned by a neural network;

said neural network for converting the electrical signals from said input means into electrical signals corresponding to at least three parameters and corresponding to the inputted electrical signals and obtained by the multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution, and for outputting the same; and

estimating means for estimating a spectral reflectance distribution or a spectral transmittance distribution by the electrical signals outputted from said neural network and by linear polynomial approximation using the eigenvectors and the mean vector which are stored in said storage means.

12. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 11, further comprising:

output means for outputting the spectral reflectance distribution or the spectral transmittance distribution estimated by said estimating means as an output signal of a predetermined form.

13. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 11, wherein said neural network

measures the spectral reflectance distribution of each a multiplicity of color chips corresponding to the at least three device values used for color reproduction with a color reproduction apparatus and each having a color whose density differs for each of the device values used for color reproduction with a color reproduction apparatus;

determines an eigenvector by performing principal-component analysis by using a covariance matrix of values in each of which the mean vector of the spectral reflectance distribution is subtracted from the spectral reflectance distribution;

determines the at least three principal components corresponding to each of sets of device values used for color reproduction with a color reproduction apparatus, including at least three device values used for color reproduction with a color reproduction apparatus, by using the determined eigenvector and the values in each of which the mean vector of the spectral reflectance distribution is subtracted from the spectral reflectance distribution; and

undergoes learning by using the sets of device values used for color reproduction with a color reproduction apparatus and the principal components such that the at least three characteristics parameters corresponding to the set of device values used for color reproduction with a color reproduction apparatus are outputted when the set of device values used for color reproduction with a color reproduction apparatus, including the at least three device values used for color reproduction with a color reproduction apparatus, is inputted.

14. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 11, wherein the set of device values used for color reproduction with a color reproduction apparatus is RGB values which are three primary colors of additive color mixture, CMY values which are three primary colors of subtractive primary colors, or CMYK values.

15. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 11, wherein said neural network is a multilayered feedforward neural network comprising three to five layers.

16. An apparatus for estimating a spectral reflectance distribution or a spectral transmittance distribution according to claim 11, wherein KL expansion is used as the linear polynomial approximation, and coefficients of KL expansion are used as the principal components.

17. A method of estimating a spectral reflectance distribution or a spectral transmittance distribution of a color, specific to a subject color-reproducing apparatus, comprising the steps of:

causing a neural network to undergo learning specifically for the color-reproducing apparatus, to output at least three principal components when a set of at least three device values used for color reproduction with a color reproduction apparatus of a color, which has a spectral reflectance distribution or a spectral transmittance distribution, is inputted into the color-reproducing apparatus, based on multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution;

correlating a set of the device values used for color reproduction with a color reproduction apparatus of a color with a spectral reflectance distribution or a spectral transmittance distribution of the color, by using a linear polynomial approximation using the principal components, eigenvectors obtained by the multivariate analysis, and a mean vector of the spectral reflectance distribution or the spectral transmittance distribution;

inputting a set of the device values used for color reproduction with a color reproduction apparatus of a subject color into the color-reproducing apparatus; and

estimating a spectral reflectance distribution or a spectral transmittance distribution of the subject color, based on the correlation between a set of the device values used for color reproduction with a color reproduction apparatus with a spectral reflectance distribution or a spectral transmittance distribution, using the neural network which has completed learning.

18. A method of reproducing a color using a color-reproducing apparatus, comprising the steps of:

causing a neural network to undergo learning specifically for the color-reproducing apparatus, to output at least three principal components when a set of at least three device values used for color reproduction with a color reproduction apparatus of a color, which has a spectral reflectance distribution or a spectral transmittance distribution, is inputted into the color-reproducing apparatus, based on multivariate analysis of the spectral reflectance distribution or the spectral transmittance distribution;

correlating a set of the device values used for color reproduction with a color reproduction apparatus of a color with a spectral reflectance distribution or a spectral transmittance distribution of the color, by using a linear polynomial approximation using the principal components, eigenvectors obtained by the multivariate analysis, and a mean vector of the spectral reflectance distribution or the spectral transmittance distribution;

inputting a set of the device values used for color reproduction with a color reproduction apparatus of a subject color into the color-reproducing apparatus;

estimating a spectral reflectance distribution or a spectral transmittance distribution of the subject color, based on the correlation between a set of the device values used for color reproduction with a color reproduction apparatus with a spectral reflectance distribution or a spectral transmittance distribution, using the neural network which has completed learning; and

reproducing a color using the color-reproducing apparatus based on the estimated spectral reflectance distribution or spectral transmittance distribution of the subject color.