Research On Characterization Method Of Printer Based On Spectrum

With the development of color science and image processing technology,digital image devices such as printers have been widely used in people's work and life,but they also bring a series of problems such as color distortion,such as the color output through the printer does not reach.The expected effect,which does not match the color on the display,etc.It can be solved by printer characterization.Printer characterization is to establish a correspondence between the printer ink recipe and the reflectance of the printed sample.The forward model is to predict the reflectance of the printed sample based on the given ink recipe,while the inverse model is to find the ink recipe based on the given reflectance of a sample.First of all,in this paper,a printer characterization model is established by using 3-ink printer data and 4-ink printer data.The 3-ink data uses 729 groups as training samples and 189 groups as test samples.CYNSN models of 3,5,and 9 levels can be established respectively.The 4-ink data uses 625 groups as training samples and 607 groups as test samples.CYNSN models of 3 and 5 levels can be established respectively,and the 4-ink data paper types also have photo paper and art paper.The three performance metrics in this paper are the root mean square error of the spectrum,the color difference,and the metamerism index.Then,based on the above data,a printer forward prediction model was established,and parameters such as the correction coefficient "n" and the number of cell levels were optimized.The reverse model belongs to the focus of printer characterization.It is also difficult.Because the reverse algorithm of the printer is irreversible,the research difficulty is greatly increased.The first step in the inverse algorithm is to determine the scope of the small cube(cell)as the core link of the optimization algorithm.so how to screen small cubes is the most important.This article first implements the predecessor's screening method,and proposes a new XYZ-based cell search algorithm and a CMY-based cell search algorithm based on the cell search algorithm.For further comparison in the measurement criteria of the optimal solution,this paper proposes two measurement criteria,one is the color difference method,and the other is the spatial distance method.By comparing the correct rate and color difference of small cube selection through experiments,this paper determines that the cell search algorithm based on XYZ is the optimal algorithm.The idea of its algorithm is to convert the spectral reflectance in the original cell search algorithm into the form of XYZ tristimulus values.Replace the spectral root mean square error in the original cell search algorithm with the color difference for comparison.Finally,the color difference method recommended in this paper is the most accurate to screen the final cell algorithm.After trying some printer reverse algorithms.This paper proposes a reverse algorithm based on tristimulus values under multiple light sources.The idea of this algorithm is to use multiple light sources and three stimulus values for dimension reduction.Each column of the matrix is based on the combination of two light sources or three light sources XYZ.It was found that according to the measure of colour differences under multi illuminants,the proposed method performs best using one illuminant(D50)for 3-ink combination,and using two illuminants(D50 and A)for 4-ink combination.The proposed methods performs not only in terms of measures of colour differences under multi illuminants,but also in terms of computational cost per iteration.Though more than 4 ink printer is available in the market,only 3 or 4 of the inks are used for a given reflectance in a real situation.How to choose the right 3 or 4 inks for a given reflectance is currently a hot research topic and this process is named as the colour separation.Hence,it is expected our proposed method has an important application for multi-ink printer characterization.