CN103530494A

CN103530494A - Subspace search algorithm in reverse CYNSN model

Info

Publication number: CN103530494A
Application number: CN201210595446.4A
Authority: CN
Inventors: 岑夏凤
Original assignee: Individual
Current assignee: Individual
Priority date: 2012-12-10
Filing date: 2012-12-10
Publication date: 2014-01-22

Abstract

The invention discloses a subspace search algorithm in a reverse CYNSN model. By calculating RRMS values between all top point spectrums and target spectrums, the top point spectrum with the minimum RRMS value is selected, all subspaces including the top point are found out afterwards and serve as candidate subspaces, and then candidate ink formulas corresponding to all the candidate subspaces are calculated through a linear iterative algorithm, spectrums corresponding to the ink formulas are predicted, and finally the candidate subspace corresponding to the minimum RRMS value between the top point spectrums and the target spectrums is selected as the optimum subspace. The algorithm is high in search speed, precision and practical value, and the optimum subspace can be acquired after calculation through few iterations.

Description

Subspace search algorithm in a kind of reverse CYNSN model

Technical field

The present invention relates to a kind of subspace search algorithm, specifically relate to the subspace search algorithm in a kind of reverse CYNSN model.

Background technology

Along with scientific and technical development, printer has obtained increasingly extensive application in daily life work, when being greatly convenient for people to Working Life, the series of problems such as cross-color have also been brought, as do not mated with the color on display through the color of printer output etc.

The characterization of equipment is to solve one of problem of the primary solution of cross-color problem.Apparatus characteristic refers to by specific mathematical model and training sample sets up the correspondent transform relation between device independent look and equipment related colour.Device independent look refers generally to the color characterizing in CIE standard color space, as CIE XYZ tristimulus values, CIELAB colorimetric parameter etc.; Equipment related colour is that the color characteristics by respective digital vision facilities determines, for printer, its equipment related colour is often referred to the consumption of each ink, i.e. ink set.Characterization model generally comprises forward and reverse model.Forward characterization model is set up from equipment related colour to the contact device independent look, and opposite feature model refers to by device independent look and predicts corresponding equipment related colour.In printer characteristic model investigation in the past, often select CIE XYZ tristimulus values or CIELAB value as device independent look, reason is to calculate relatively simple, but because CIEXYZ or CIELAB value do not comprise the information of all colours, can cause occurring the problem of metamerism, be that master sample and reproduction sample its color under a kind of lighting condition are mated mutually, but under another kind of lighting condition unmatched phenomenon.Because spectroscopic data has recorded all information of object color, therefore the characterization model based on spectrum can solve the problem of metamerism well.

CYNSN (Cellular Yule-Nielsen Spectral Neugebauer, the spectrum Neugebauer that subspace formula Yule-Nielsen revises) model is the printer spectrum characterization model of commonly using the most, it is the improved model of basic Neugebauer model, comprises forward and reverse model.The opposite feature model of this model (based target Forecast of Spectra ink set) always is the focus of correlative study.In CYNSN model, whole ink space is divided into a plurality of subspaces, so in reverse CYNSN model, for a certain target optical spectrum, first need to find the subspace at target optical spectrum place, then according to linear iterative algorithm, this target optical spectrum is recorded to ink set in advance.Up to the present, seldom there is simple, effective subspace search algorithm to find fast the affiliated subspace of optimum target optical spectrum.

Summary of the invention

The present invention, in order to solve the problem described in background technology, discloses the subspace search algorithm in a kind of reverse CYNSN model.Its concrete steps are as follows:

1) choose n ink of printer, each ink is divided into m section from 0 (without ink) to 1 (maximum quantity of ink), altogether whole ink space is divided into m ⁿsub spaces, each subspace comprises 2 ⁿindividual summit, (m+1) contained in whole ink space ⁿindividual summit; In general, for obtaining higher model accuracy, the value of m should be more than or equal to 6;

2) print all (m+1) ⁿindividual summit is as training sample, adopt afterwards spectrophotometer to measure to obtain corresponding spectrum to all training samples, according to the establishment of spectrum forward CYNSN model of all training samples that record, the foundation of forward CYNSN model can be referring to journal of writings (the Optimizationof the spectral Neugebauer model for printer characterization.Journal of ElectronicImaging of R.Balasubramanian, 1999, the 8th the 2nd phase of volume, 156-166);

3) for a certain target optical spectrum, the RRMS of calculating and each summit spectrum, computing formula is as follows:

RRMS = \sqrt{\frac{Σ_{λ = λ 1}^{λ 2} {(R_{1} (λ) - R_{2} (λ))}^{2}}{N}}

In formula, R ₁(λ) and R ₂(λ) be the spectrum on target sample and a certain summit, λ 1 and λ 2 are minimum value and the maximal value within the scope of measure spectrum, and N is the number of samples within the scope of measure spectrum.

4) select that summit minimum with target optical spectrum RRMS value, find out afterwards all subspaces that comprise this summit, be called candidate subspace, be assumed to t candidate subspace;

5) in each candidate subspace, with linear iterative algorithm, calculate the ink set of target optical spectrum, obtain altogether t candidate's ink set, linear iterative algorithm adopts UG algorithm (P.Urban and R.R.Grigat, Spectralbased color separation using linear regression iteration.Color Research Application, 2006, the 31 phase the 3rd volumes, 229-238);

6) for each candidate's ink set, utilize its corresponding spectrum of forward CYNSN model prediction, calculate the RRMS value of this prediction spectrum and target optical spectrum simultaneously;

7) finally selecting minimum corresponding that candidate subspace of RRMS value in all prediction spectrum is optimal subspace, the final ink set that the ink set calculating in this candidate subspace is target optical spectrum.

The present invention is by selecting to find all subspaces that comprise this summit with that summit of target optical spectrum RRMS minimum, thereby greatly dwindle the hunting zone of subspace, accelerate the speed of algorithm, then utilize linear iterative algorithm to calculate candidate's ink set corresponding to all candidates subspace, and predict the spectrum that it is corresponding, finally select corresponding that candidate subspace of RRMS value minimum between prediction spectrum and target optical spectrum as optimal subspace.The method search speed is fast, precision is high, only needs less iterations just can calculate optimal subspace, can obviously improve the solution efficiency of the more reverse CYNSN model in subspace.

Accompanying drawing explanation

Fig. 1 is the subspace search algorithm flow chart in a kind of reverse CYNSN model;

Fig. 2 is the target optical spectrum in embodiment;

Fig. 3 is the RRMS value of prediction spectrum and the target optical spectrum of 8 candidate's ink sets in embodiment;

Fig. 4 is prediction spectrum that in embodiment, final ink set is corresponding and the comparison of target optical spectrum.

Embodiment

Traditional three black CMY (green grass or young crops, product, the Huang) printer of take is example, and the embodiment of the subspace search algorithm in above-mentioned a kind of reverse CYNSN model is set forth.As shown in Figure 1, its concrete steps are as follows:

1) so choose 3 inks of printer, each ink is evenly divided into 7 sections from 0 (without ink) to 1 (maximum quantity of ink), altogether whole ink space is divided into 7 ³=343 sub spaces, each subspace comprises 2 ³=8 summits, whole ink space contains 8 ³=512 summits;

2) print all 512 summits as training sample, adopt afterwards spectrophotometer to measure to obtain corresponding spectrum to all training samples, according to the establishment of spectrum forward CYNSN model (Yule-Nielsen modified value is made as 2.5) of all training samples that record; Wherein, spectrophotometric measurement range is made as 400nm to 700nm, take 10nm as interval;

3) for a certain target optical spectrum, see Fig. 2, the RRMS of calculating and each summit spectrum, computing formula is as follows:

RRMS = \sqrt{\frac{Σ_{λ = λ 1}^{λ 2} {(R_{1} (λ) - R_{2} (λ))}^{2}}{N}}

In formula, R ₁(λ) and R ₂(λ) be the spectrum on target sample and a certain summit, λ 1 and λ 2 are minimum value 400nm and the maximal value 700nm within the scope of measure spectrum, and N=31 is the number of samples within the scope of measure spectrum.

4) select that summit minimum with target optical spectrum RRMS value, find out afterwards all subspaces that comprise this summit, be called candidate subspace, in this embodiment t=8 candidate subspace; Wherein, the value of t is relevant with the position in whole ink space on this summit, t=1 or 4 or 8;

5) in each candidate subspace, with linear iterative algorithm, calculate the ink set of target optical spectrum, obtain altogether t candidate's ink set, linear iterative algorithm adopts UG algorithm;

6) for each candidate's ink set, utilize its corresponding spectrum of forward CYNSN model prediction, calculate the RRMS value of this prediction spectrum and target optical spectrum simultaneously, see Fig. 3;

7) finally selecting minimum corresponding that candidate subspace of RRMS value in all prediction spectrum is optimal subspace, the final ink set that the ink set calculating in this candidate subspace is target optical spectrum, this final ink set is C=0.369, M=0.098, Y=0.122, the prediction spectrum of its correspondence and target optical spectrum are as shown in Figure 4.

Claims

1. a subspace search algorithm in reverse CYNSN model, is characterized in that comprising the following steps:

1) choose n ink of printer, each ink is divided into m section from 0 (without ink) to 1 (maximum quantity of ink), altogether whole ink space is divided into m ⁿsub spaces, each subspace comprises 2 ⁿindividual summit, (m+1) contained in whole ink space ⁿindividual summit;

2) print all (m+1) ⁿindividual summit, as training sample, adopts spectrophotometer to measure to obtain corresponding spectrum to all training samples, according to the establishment of spectrum forward CYNSN model of all training samples that record afterwards;

RRMS = \sqrt{\frac{Σ_{λ = λ 1}^{λ 2} {(R_{1} (λ) - R_{2} (λ))}^{2}}{N}}

2. the subspace search algorithm in a kind of reverse CYNSN model according to claim 1, it is characterized in that described step 1) in choose n ink of printer, each ink is divided into m section from 0 (without ink) to 1 (maximum quantity of ink), altogether whole ink space is divided into m ⁿsub spaces.In general, for obtaining higher model accuracy, the value of m should be more than or equal to 6.

3. the subspace search algorithm in a kind of reverse CYNSN model according to claim 1,, it is characterized in that described step 2) in print all (m+1) ⁿindividual summit, as training sample, adopts spectrophotometer to measure to obtain corresponding spectrum to all training samples, according to the establishment of spectrum forward CYNSN model of all training samples that record afterwards.The foundation of forward CYNSN model can adopt the method for R.Balasubramanian.

4. the subspace search algorithm in a kind of reverse CYNSN model according to claim 1, it is characterized in that described step 4) in select that summit with target optical spectrum RRMS value minimum, find out afterwards all subspaces that comprise this summit, be called candidate subspace.

5. the subspace search algorithm in a kind of reverse CYNSN model according to claim 1, it is characterized in that described step 5) in each candidate subspace, with linear iterative algorithm, calculate the ink set of target optical spectrum, linear iterative algorithm can adopt UG algorithm.

6. the subspace search algorithm in a kind of reverse CYNSN model according to claim 1, it is characterized in that described step 7) in final to select corresponding that candidate subspace of RRMS value minimum in all prediction spectrum be optimal subspace, the final ink set that the ink set calculating in this candidate subspace is target optical spectrum.