How to perform color data conversion

Abstract

To provide a method of performing color space conversion by using a conversion rule in a small amount of interpolation points.SOLUTION: In a method, m nominal value datasets (Z1) in a color space of a supply destination which have been verified are determined about a set of color data (called Q1) from n colors in a color space of a supply source. A copy of Q1 is created as a dataset Q2. The set of color data Q2 is converted by using a conversion rule TRV which is provided. Thus, another dataset Z2 of m values in the color space of a supply destination is created. The dataset Q2 is changed in an optimization procedure and is returned to Z2 until a difference between Z1 and Z2 becomes minimum. Thus, a difference Dq between an original dataset Q1 and an optimized dataset Q2 and a difference Dz between an optimized result Z2 and the dataset Z1 of the supply destination are applicable to conversion of color data into a color space of the supply destination of m values from a color space of the supply source of n colors that uses the conversion rule TRV.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for performing color data conversion from a source color space having n colors to a destination color space having m values using a conversion rule TRV.

Background Art

[0002] In the printing industry, color conversion may be required to create a dataset of a color printing product to be printed.

[0003] The starting point is a digital layout. The digital layout may include, for example, color images captured by photography or scanning, illustrations, documents, etc. Such a layout is a template that is reproduced as it is in the form of a printed product. Such a layout is designed or developed in an arbitrary format and, from the perspective of color, represents what is desired according to the intended purpose. Usually, a layout includes several layout objects. Therefore, each layout object has a uniquely identifiable color appearance. As is well known, in the case of a printed product, for example, by simply printing screened printing inks on each other, each color makes a certain contribution through its effect, and the final result is created by the interaction of different color layers, so that a color appearance can be created by mixing only a small number of colors.

[0004] The color appearance of each layout object is defined by color values ​​(image pixels, outline colors, or fill colors) that represent the data source. In the case of image pixels, there are typically millions of pixels, each with its own color value. These source color values ​​always refer to the supplier's color space, which has a specific number of colors. This means that each object in the layout is assigned to a color space typically associated with the device, which can be a specific RGB color space such as sRGB or AdobeRGB, a CMYK color space, a black-only (grayscale) color space, or other print color spaces. The object's color values ​​are quantified as RGB values, CMYK or grayscale values ​​(0% to 100% color), or other color values, where the total number of device colors in the supplier's color space is denoted by n.

[0005] Color spaces are typically represented by ICC color profiles (ISO 15076-1:2010) and by layouts in PDF and PDF / X files (ISO 15930 series) that contain or reference these ICC color profiles.

[0006] Layouts are reproduced for vastly different reasons and purposes. A wide variety of printing processes and presses can be used for reproduction. This has the potential to apply a specific number of colors to the carrier material in a specific way; that is, the colors used by the devices in the printing system. These colors define or form the basis of the target color space and therefore include all possible combinations of colors used by the devices in the printing system. This is referred to as the m device colors of the target color space. Often, conversion to device-independent values ​​such as Lab or spectral color (see below) is also necessary when verifying post-print measurement results or providing a device-independent database for display on any device. This is why the term "m values ​​of the target color space" is commonly used.

[0007] Generally, the color spaces of the source and destination do not match. Standardized color spaces are typically used as the source color space during layout design, but printing is performed using specific output systems with their own unique color behavior. Therefore, for each halftone dot in the layout, a predetermined combination of n device colors from the source color space needs to be converted to a combination of m device colors from the destination color space required for printing. The goal here is to produce a print product that is as close as possible to the original, i.e., the digital layout, and ideally, to reproduce the same appearance to the human eye.

[0008] A halftone point, or pixel, is any point on a layout that results from the specified rasterization of a layout object. This relates, for example, to the so-called image resolution for creating print forms. If a grid with i rows and j columns is placed on the layout, the resolution can be expressed as i × j pixels. Individual pixels can be identified from one another by their color structure.

[0009] Color conversion is performed when converting layout data into printable data with the desired print color set. This results in, for example, uniform and authentic CMYK data for offset printing. Color conversion is also performed to bind and simulate the color appearance of this print data before printing. The color appearance is usually represented either device-independent in the CIELAB color space (abbreviated as Lab) or using spectral reflectance. Print data with the assigned color space is converted into device-independent color data. These are displayed on a calibrated screen or a special proofing printing system (a so-called "proof," usually an inkjet system). In connection with this, there is the task of adapting print data to other printing processes while preserving the impression of the original colors as much as possible; that is, the task of generating new print data for other inks from existing print data through conversion.

[0010] Generally, for typical printing processes with four colors or less, these conversion processes are largely established and standardized by the corresponding data systems.

[0011] However, in packaging printing, for example, four or more printing colors are generally used. Traditionally, brand colors are added as spot colors, and images are constructed with product-specific colors (for example, a light gray printing color is added for dairy products, and various shades of brown are added for chocolate products). In addition, white printing ink is often added to transparent film packaging. Each ink used occupies a printing unit of the printing press and requires its own printing form in ink transfer. Colors are applied sequentially, and in principle, all colors can be combined with each other (overprintable). The wholeness in separate printing forms leads to the process of creating print data, or consequently, the term "separation."

[0012] Machines with 7 to 10 ink units are not uncommon in flexographic, offset lithography, and gravure printing. Designers of branded products can choose printing inks from thousands of spot colors, including over 2,000 hues in the PANTONE (trademarked) system. These inks are manufactured and supplied by ink manufacturers. In so-called digital printing, inkjet and toner-based printing presses are used, often containing 6 to 8 ink units, but operating with fixed ink sets or changing them is too technically complex because inkjet inks and color toners require highly specific, machine-dependent processing characteristics. Typical color sets include CMYK, complemented by orange or red, green, blue or purple, and white.

[0013] Currently, while digital printing processes are slower than traditional printing processes, they significantly reduce setup effort because they eliminate the need to create print forms or change colors. This allows for faster, more economical printing and enables personalized printing. Therefore, even if there is a demand for rapid printing of conventionally produced packages or for personalization as needed (execution 1), the color appearance must be as identical as possible to previously printed products manufactured with individual spot colors. This is to allow different batches of products to be displayed on shelves. For this purpose, print data usable in traditional printing processes is converted from its source color space to the destination color space representing the digital printing process. In the case of extremely modified or personalized printing, data from different sources must be converted for each copy, which becomes a bottleneck.

[0014] Therefore, the color conversion described at the beginning converts from the source color space Q (e.g., having device color) to another source color space Z that has m values ​​(e.g., device color or three CIELAB values ​​L, a, b). This is done by applying a conversion rule TRV. Such rules refer to calculation formulas, conversion models, or conversion tables that have already been created based on TRV. High throughput is important because it needs to be applied frequently per pixel (millions of pixels per second). To achieve this, interpolation is performed using a predefined table such as an ICC industry standard. The table is constructed assuming n combinations of color values. Each of the n input colors can take values ​​between 0% and 100% ink coverage. Not all combinations can be listed on the table, but each input color is sampled at a specified resolution. An n-dimensional grid of squares containing all combinations of sampled levels is obtained. Between them, n-dimensional interpolation is applied. The accuracy depends on this sampling. A typical sample is n=3 (RGB / Lab), which has 33 levels (33 3(approximately 36,000 table entries), and for n=4 (CMYK), there are 17 grayscale levels (6.25% increments, 17 4 (Approximately 84,000 table entries). The number of gradations represents the base in the exponent, and the number of inks n is the exponent. The number of gradations is the base of the power, and the number of inks n is the exponent. The generation of conversion tables is a method familiar to those skilled in the art. They are based, for example, on spectroscopic measurements of printed inks, calculations for estimation and / or interpolation, etc.

[0015] The size of the color table increases exponentially with the number of input colors n. Due to limited memory space, it becomes necessary to coarsely sample for practical reasons. For a typical n=7, only 6 or 7 levels are used. One example is the widely used open-source color management system LittleCMS, which uses 7 levels (16.67% increments). 7 This amounts to approximately 820,000 table entries. Therefore, the table structure has the problem that the larger the number of inks, the larger the table becomes, and the fewer gradients can be stored. Users expect the same level of precision they are used to with CMYK to be achieved with CMYK and additional colors. Therefore, 7 levels is quite coarse.

[0016] A known solution approach is disclosed in Patent Document 1. According to this, an assignment table with four or more input color components is divided into several assignment tables with a maximum of four input color components each. This method aims to solve the problem of assignment tables being too large. In the prior art, even with 7-color printing, typically only four or fewer colors are printed overlapping at a single point on the printed sheet. Accordingly, a concrete proposal was made to solve the above problem by essentially reducing it to a 4-component table. For this reason, the reduction to a 4-component table is also being promoted. To make it more reliable, additional processes such as gray component substitution (GCR) and color component substitution (CCR) are applied when creating the separation table. Similar prior art can be found in Non-Patent Document 1 and Patent Document 2.

[0017] The known solution approach uses a model or a combination of a model and a table. A model is typically a mathematically based transformation rule. In this way, the Lab color values ​​of overprints can be estimated from individual print colors. This is done, for example, when displaying layouts or PDF files on a screen, and is fast but lacks accuracy. Further development of ICC, such as iccMAX, proposes either spectrally offsetting monochromatic colors or using a table for the CMYK portion and changing the value for each additional color. In either case, adding individual colors is highly inaccurate because it does not, or does not adequately, consider the interactions of the overprint colors in which these colors are involved.

[0018] While more accurate models exist to better calculate the overprint behavior of color mixing, these models are not suitable for processing at speeds fast enough to be applied to conversions of millions of pixels. Furthermore, the models provide Lab values ​​or spectra, but not device color values, making them unsuitable for direct conversions between devices ("device links"). This is because the color values ​​of the output device often follow special separation rules (see below), and typically require specific processing to improve quality. Nevertheless, conversion to device color has become very common.

[0019] Color transformations are often applied to higher-dimensional spaces. These occur on both the input and output sides.

[0020] On the input side, for example, it can be a combination of colors for the printing process. Traditionally, cyan, magenta, yellow, and black were the standard colors for multicolor printing. In recent years, additional colors have been used to expand the chromaticity. In particular, efforts are underway to standardize 7-color printing by adding orange or red, or green, blue, or purple (known as ECG printing or extended color gamut). Even more colors are used in specialized applications such as packaging printing and banknote printing.

[0021] The output side contains m values, such as CIELAB color values ​​(3 components), a combination of print colors for the target system, and spectral reflectance coefficients (values ​​in the wavelength range, for example, m=31 values ​​in the 10nm band from 400 to 700nm).

[0022] A common application is predicting color effects achieved through the spectral reflectance (output side) of proportionally overprinted colors (input side) in industrial printing processes. Only when such printing characteristics are known can further conversions be performed to consistently match colors from one printing process to another. Color conversions are based on tables or models, as described above.

[0023] Models can be empirical or mathematical, or mechanical and physical, for example, to analyze color behavior. While both modeling approaches involve tuning model parameters, there is usually a residual deviation between the model and reality, which needs to be reduced. Therefore, a rigid model following internal laws needs to be adapted to independent source data.

[0024] Tables with multidimensional interpolation offer greater flexibility because entries can be freely adjusted. The problem here is that as the number of input dimensions (print colors) n increases, the storage space for color combinations increases exponentially, making detailed storage impossible. One recommended solution is to store only combinations that occur at the same location in overprints. These are typically only four overprint colors. Therefore, several 4-dimensional tables suffice rather than a large n-dimensional table. Nevertheless, in a color processing chain, there may be more than four color components that need to be converted. The proposal here is to estimate the output value by assembling multiple contributions from existing tables, as detailed elsewhere. For example, using the largest 4-dimensional contribution, determine their color effects from other tables in which they occur, and add missing colors by applying them to a large 4-dimensional output value. This estimation procedure is ultimately a simple model, which usually deviates from reality. However, it cannot be directly adjusted because there are no table locations beyond 4 dimensions.

[0025] In this regard, it is desirable to correct the calculation results of the table or model to a predetermined nominal value. The correction should not only have a selective effect, but also gently adapt to the overall behavior of the surrounding environment.

[0026] The latest technologies include correction methods that compensate for the difference between target values ​​and actual values. Patent Document 3 describes a combination of colors printed from the color components of a printer source color space, with expected reference values ​​in a device-independent destination color space, Lab (m=3) (here, CMYK, n=4). Using a specific printing device, the color combination is printed, the actual Lab values ​​are measured, and the deviation is determined as a difference. Based on this difference, the inverse conversion table is corrected from the absolute Lab color space to the printer color space. This improves the conversion from Lab color values, which are not specific to a particular device, so that they are output accurately to a specific device. However, the original conversion from the source to the destination color space is not corrected to the actual values.

[0027] Patent Document 4 mentions a specific printing device. After determining the need for calibration, a difference profile is generated, which is not described in detail, "synthesized" with a standard profile, and color-corrected print data is generated, and printed by the synthesized profile. The goal here is also color correction due to changes and differences in individual devices. Here too, the original conversion from the source color space to the destination color space is not corrected to the actual values.

Prior Art Documents

Patent Documents

[0028]

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Non-Patent Documents

[0029]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0030] In view of the above-mentioned prior art, the present invention is based on the problem of improving a method for performing color space conversion using a conversion rule in a way that can correct the results of a table or model calculation (n input values ​​corresponding to m output values, hereinafter referred to as the core model) to a given nominal value. [Means for solving the problem]

[0031] A technical solution to this problem is provided by a method having the features described in claim 1 of the claims. Further advantages and features are revealed in the dependent claims.

[0032] This invention is based on the additional use of a correction process.

[0033] In this invention, corrections are also stored in a multidimensional table, but managed with a very small number of interpolation points, in the best case only two points (0 and 100% of the input color), including only the necessary number of dimensions and not necessarily all n input colors. Some of these corrections can be applied sequentially to n different, reliably overlapping subspaces of colors. Due to the small number of interpolation points, the interpolation corrections automatically become flexible and broad.

[0034] Therefore, the present invention proposes a method for performing the conversion of color data from a source color space with n colors to a destination color space with m values, using the conversion rule TRV. a ) The combination of components z(1), z(2), ..., z(m) in the color space of the source. A value dataset Z1 containing the following is assigned to a color dataset Q1. Color data of n colors in the source color space having color components q(1), q(2), ..., q(n) Tase The process of selecting Q1, b) Initially, a set that is the same as Q1, and has n colors from the source color space, with color components q'(1), q'(2), ..., q'(n). color data set Q 2 The process of production, c) Using the transformation rule TRV, The aforementioned color data set Q 2 The process involves converting and generating an actual value dataset Z2 of m values ​​of z'(1), z'(2), ..., z'(m) in the color space of the source, d) The aforementioned Color data set Q Based on the n-dimensional optimization procedure change This is a process that The process involves repeating step c) to minimize the difference between the nominal value dataset Z1 and the actual value dataset Z2 in least squares error, and applying an n-dimensional optimization procedure so that Q2 ultimately contains the optimized color components q'(1), q'(2), ..., q'(n), e) Optimized The aforementioned color Each color part of n colors in dataset Q2 and The aforementioned color The difference Dq between each color portion of dataset Q1. (j) to Dq (1)=q'(1)-q(1), Dq(2)=q'(2)-q(2),..., Dq(n)=q'(n)-q(n) by The calculation process, f) value Each component of dataset Z1 and value The difference Dz from each component of dataset Z2. (j) to Dz (1)=z(1)-z'(1), Dz(2)=z(2)-z'(2),..., Dz(m)=z(m)-z'(m) by The calculation process, g) When converting the dataset, for each value from i=1 to n, the color data q(i) of n colors in the source color space is used. The aforementioned By adding the difference Dq(i), when converting a set of color data from the source color space of n colors to m values ​​in the destination color space, The aforementioned The difference Dq(1), Dq(2), ..., Dq(n) and The aforementioned difference The process involves applying Dz(1), Dz(2), ..., Dz(m), and the color data q'(i) thus converted is then converted by the conversion rule TRV to the value of m from j=1 in the supplied color space. It is a single value. It is converted to z'(j), Each component of z'(j) above, handle The aforementioned Difference Dz(j ) each component z'(j) butThe additional steps, It is characterized by including.

[0035] To perform correction, it is first necessary to select an appropriate and reliable method for adjusting the output value. Simple possibilities include, for example, a difference that is added or a factor that is multiplied, or other mathematical functions of the value and the correction value. This choice determines the type and application of the table entry. Differences are suitable when the correction in the source color space is independent of the magnitude of the output value, and have the advantage of more restricting the corrected value. Factors are suitable when the nature of the correction is more relative, i.e., dependent on the value, for example, in the case of spectral reflectance, where the model error is often understood as a deviation of transmittance, and ultimately it is a multiplicative process. If no correction is needed, an invalid entry is used (e.g., 0 for addition, 1 for multiplication). Factors can be easily traced back to differences by performing the correction in logarithmic space.

[0036] This invention assumes that for a set of color data (referred to as Q1) from n colors in a source color space, there exists a protected, i.e., verified nominal value dataset of m values ​​in a destination color space. This dataset represents a kind of reference and is denoted as Z1. The verified dataset can be obtained from intrinsic processing results, for example, by printing n color components and spectrally measuring the results. Alternatively, it can be obtained, for example, from an elaborate model calculation for n color components point by point. If the destination color space represents device-independent Lab values ​​or spectra, a spectral overprint model can be used for this purpose. If the destination color space represents m device colors, a color separation algorithm can be used to calculate a suitable destination color structure with m device color components in the destination color space, taking printing rules into account.

[0037] A second set of color data, Q2, is created as a copy of the color data set Q1 and can then be modified independently. Q2 is transformed using the desired transformation rules (such as a transformation table). This generates another dataset of m values ​​in the target color space. This is called Z2 and represents the actual values ​​of the transformation.

[0038] The difference between two datasets, Z2 (the transformed result of Q2) and Z1 (the nominal value dataset), is calculated. This difference can be applied to the transformation of color data from a source color space with n colors to a destination color space with m values, using the transformation rule TRV. For example, these differences are recorded in a correction table.

[0039] According to the present invention, data within a set of color data Q2 containing n colors from the source color space is modified and transformed using a provided transformation rule. This generates a new Z2. This correction of Q2 and transformation to Z2 is performed until the difference Dz between two datasets Z1 and Z2 is minimized as the least squares error. For this optimization task, n-dimensional nonlinear optimization methods well known in numerical mathematics, such as the Levenberg-Marquardt algorithm, are suitable. This ultimately generates a set of optimized color data Q2 from the n colors of the source color space, and the difference Dq of the corrected data Q2 that produces the smallest difference from the unmodified data Q1 can be found.

[0040] These differences Dq and Dz can be applied to the conversion of color data from a source color space with n colors to a destination color space with m values, using the conversion rule TRV.

[0041] When multiple correction steps affecting overlapping subspaces are performed sequentially, the results typically differ depending on the sequence. Therefore, the correction values ​​of later steps also depend on previously performed corrections. Accordingly, according to the present invention, it is advantageous to generate a correction table from nominal values, core model TRV, and sequence as follows.

[0042] The available destination output values ​​are given in the form of one or more coarse-level multidimensional tables, which, if necessary, cover up to n colors in a subspace larger than the stored fine-level (e.g., k-dimensional) tables or model calculations (u colors). A correction chain is constructed in the order of these tables, with the first one being empty.

[0043] In the initial setpoint table, the core model is invoked for all entries (as the input dataset), the results are compared to each setpoint of all m components, the input values ​​are optimized as needed, the required correction values ​​for the input are calculated based on the optimization, and the output is calculated based on the remaining difference to the setpoint. Two corrections are input into two new correction tables. These tables are placed around the core model as additional correction transformations on the input and output sides.

[0044] Next, the following setpoint table is processed in a similar manner. In the resulting second pair of correction tables, the effect of the first table is taken into account, and if there are duplicate entries, no incorrect multiple corrections are made. Of course, duplicate, redundant setpoint entries must contain the same output value. Therefore, the setpoint tables themselves must be consistent.

[0045] Ultimately, a procedure is available to perform soft adjustments to nominal values ​​using a relatively small number of grid-based control points by calling a core model and a series of upstream and downstream corrections.

[0046] In a sense, it is a collection of concentric shells around the core model, each shell consisting of input-side adjustments and output-side adjustments, both defined by the same grid of input values. That is, the sequence of table entries for a table pair is the same, with each entry having both n input correction values ​​and m output correction values, and the input correction is applied first when the input values ​​are transformed, followed by the next innermost shell (or inside the core model) being called, and the output correction is applied after the return.

[0047] Depending on the application, it is reasonable to apply only one or both of the differences Dz(i) and Dq(i) in step g). Generally, using both difference tables means that the input dataset is first accurately converted to nominal values, but not only is a simple interpolation of the deviation performed, systematic deviations are already compensated for on the input side. Furthermore, by expanding the possibilities for correction and improving the correction, it offers considerable advantages compared to state-of-the-art techniques.

[0048] To reduce the number of steps, it is proposed that in step d), only the non-zero color components of n colors are changed.

[0049] The method according to the present invention is carried out on a computer unit by control software in accordance with the proposed invention. Here, the computer unit comprises an input unit that provides digital color data of a project, an output unit that outputs converted color data, and a memory that stores a conversion table. Here, the color data of the destination color space is generated by the control software for inputting the project's color data using the conversion table and is provided as a dataset.

[0050] A more formal and generalized explanation of the process is as follows:

[0051] Let Rn be the n-dimensional space of input vectors, and Rm be the space of m-component output vectors. For a point P from the subspace Ru of the input space Rn, we assume that its nominal value T from u≦n-dimensional Rm is known. Correction shells are placed around the core model M (Rn->Rm). A correction shell typically consists of adjustment A of the model's previous input values ​​(Ru->Ru) and coupled adjustment B of the model's later output values ​​(Ru->Rm). The corrected model (A->M->B) is still the mapping n->m. Further shells can be placed around the corrected model. A or B can be omitted. Formally, they can be considered the same mapping. In implementation, each step is skipped. A and B are determined by processing all known points P as follows:

[0052] An input point P from Ru is associated with a setpoint T from Rm. The model provides a model output value V for this input vector (missing nu-dimensional entries are filled with zeros).

[0053] Image A serves to prepare the input values. Next, for each point P, a corrected input value P' is found that is close to the model output value V' T. A commonly used method is the Levenberg-Marquardt method, which is a multidimensional nonlinear optimization of the input parameter P, and constraints are used as needed, such as the effective domain of the definition. Thus, mapping A consists of pairs (P, P'). This step is performed before determining B.

[0054] When mapping B is used, the model (pre-corrected by A if necessary) is referenced for each point P. That is, if image A is used, P -> P' is mapped to it, and P' is given to the model. If mapping A is not used, P is given directly to the model. The model provides a model output value V with m components. The deviation from the nominal value T can be seen, for example, as the difference D = TV per component or the factor F = T / V per component, so the model is corrected to the nominal value T by V + D = T or V × F = T. Mapping B then consists of pairs (P, D) or (P, F) and provides the correction values ​​to be applied to V.

[0055] Mappings A and B need to be interpolatable so that these point-by-point corrections can gently affect the neighborhoods of known points. Here, topological methods such as "natural neighborhood interpolation" or "thin-plate splines" are possible and can be constructed at any point. However, the most efficient and easily controllable method is a regular orthogonal u-dimensional grid table established in color transformations such as ICC profiles, or a collection of combinations thereof.

[0056] The following describes and illustrates an example of the application.

[0057] Example 1: The goal is to convert seven print color values ​​into spectral reflectance values.

[0058] For this purpose, there are sophisticated (physical) predictive models that can calculate reflectance values ​​for any combination of overprints. However, these tables are too large to be applied to millions of pixels in a print file, and inputting predictions into detailed 7-dimensional tables is too costly.

[0059] Therefore, this predictive model sets up only a collection of detail tables for all 4-dimensional subspaces to cover the important color combinations that typically occur in printing. Furthermore, there is a component that can quickly (but roughly) estimate 5-7 color queries from known 4-dimensional tables. The results of the collection and estimated components form Model M.

[0060] Since this estimate is a kind of extrapolation (to additional dimensions not tabulated), it is advantageous to control it selectively. Now, as Image B, we can add a collection of roughly graded 5-dimensional subspace tables. Here again, we have estimated components for 6-color or 7-color queries. For each entry P, we get a correction D or F from the estimated result V from M for the target value T from the predictive model. This has already improved the 5-dimensional queries. (For 4-dimensional entries, we can directly set correction D=0 or F=1).

[0061] As a further correction shell, only the corners (0 or 100%), i.e., 2 7 Further images B2 can be added using a 7-dimensional table containing only entries. Some of these can be appropriately computed using a predictive model. Thus, free extrapolation is fully understood, providing reasonable and controlled results for all queries.

[0062] Example 2 (Figures 1 and 2): The objective is to fit the physical prediction model to the measured values ​​of the test chart.

[0063] Physical prediction models describe the optical behavior of colors when they are printed or mixed in layers. Printed substrates are used as the basis. The properties of individual inks are characterized, for example, by the effective halftone dot area ratio, layer thickness, and transmission and scattering spectra. For this purpose, color patches consisting only of gradients of individual printing inks are used. Ink transfer from ink to paper, and ink transfer from ink to already printed ink, are process-dependent. This so-called ink reception behavior is typically determined by ink patches consisting of pairs of inks applied across the entire surface. This means that the printed image is essentially physically captured. This allows for the calculation of the spectral reflectance of any color combination.

[0064] However, this model is merely a simplified representation of reality. Therefore, we consider the case where additional color patches are used to softly fit the model to actual measurements. For this purpose, there are established test charts such as the ECI2002 test chart for the CMYK printing process. This test chart includes all CMYK combinations, in particular, for tone values ​​of 0, 40, or 100%. These three 4 Using 81 fields, it is possible to match image A with image B.

[0065] Image A corresponds to screen halftone interactions that are not properly represented by the model. For example, if a 40% cyan overprint is followed by a 40% magenta overprint, the model may show a more reddish color than it actually is because the magenta halftone dots do not adhere as well to the cyan halftone dots as they would to a full solid ink transfer. Then, the input dot (40,40,0,0) maps to the modified dot (40,36,0,0). In a physical interpretation, magenta shifts slightly less than 10% to cyan in the halftone.

[0066] Image B shows the residual error of the model after this preliminary adjustment, so the entire model is (0, 40, 100). 4Accurately reproduces measurements on the grid.

[0067] This bends the model at key points. The ECI2002 test chart also includes finer CMYK combinations, but these are not perfect in the direction of black, so individual nominal values ​​are missing. Next, we can place another correction shell around the model using the CMY(0,20,40,70,100) and K(0,20,40,80,100) combination. The previously corrected (0,40,100) combination gets zero correction in the outer shell. At the new level, the inner shell (0,40,100) is also significantly bent in the midtones by the interpolation of the correction, so only relatively small correction is needed. The missing nominal values ​​leave "gaps" in correction tables A and B, which are gently filled by the interpolation of the set neighborhoods. Furthermore, the third shell can be filled with the finest gradations of CMY (0, 10, 20, 30, 40, 55, 70, 85, 100) and K (0, 10, 20, 40, 60, 80, 100), accurately reproducing more than 90% of the color patches in the ECI2002 test chart.

[0068] Finally, we will explain a concrete example using a subspace table. This subspace table stores the correct values ​​for the destinations of input data with a maximum of k(4) color components, and the estimation procedure provides estimated values ​​for the destinations of input data with k color components (maximum n).

[0069] Here, the corrected transformation is shown, represented by a subspace table. The transformation provides m values ​​as "results" for input color data from n color components. This can be achieved by interpolation for any input (a combination of n colors with 0-100% shares).

[0070] For a given set of inputs, correct results exist. These are correct, verified results that match the results generated by the transformation. These results, for example, yield good results when the input color data is printed and measured, or when transformed using another model.

[0071] The following is a concrete example of calculating a subspace table when k=1 (only one color) and the total number of colors is n=2. The color space supplied is Lab, meaning m=3 values ​​(lightness value, redness value, yellowness value).

[0072] Subspace Table 1: Yellow Printable 0% Yellow 9500 (Very bright, non-reddish, non-yellowish, unprinted paper) 100% Yellow 900100 (Bright, non-reddish, very yellowish)

[0073] Subspace Table 2: Magenta Ink 0% Magenta 9500 (Very bright, non-reddish, non-yellowish, unprinted paper) 100% Magenta 55700 (Medium brightness, very reddish, not yellowish)

[0074] Furthermore, the correct values ​​for the supply destination are also known, given the input of 100% yellow + 100% magenta. 100% Yellow + 100% Magenta 546562 (depending on model or pressure + measurement)

[0075] Through conversion, 100% Yellow + 100% Magenta 526764

[0076] The difference from a set point is 100% Yellow + 100% Magenta +2 -2 -2 (A bright, reddish-yellowish color)

[0077] The optimization process reduced the yellow and magenta color components. The results may be illustrative. 98% Yellow + 96% Magenta 536462

[0078] The difference with the supplier's value will be minimized, but it will not be zero. Less magenta may result in a brighter color, but the redness will be further reduced, resulting in a value that is too low. 98% Yellow + 96% Magenta -1 -1 0 (Still too dark, not enough red)

[0079] The difference between the optimized input values ​​is assigned to the original input values. 100% Yellow + 100% Magenta - 2% Yellow - 4% Magenta

[0080] By applying this difference to the result, the transformation can be corrected for this input.

[0081] Another input, for example, 50% yellow + 50% magenta, is far removed from known input values. The target may require smaller corrections, which may attenuate with distance, for example.

[0082] The two tables from the input set, and the difference between their inputs and the difference between their values ​​at the destination, are shown in Table 1.

[0083] [Table 1]

[0084] Interpolation is performed within this range. 50% 50% -1 / 2 -1 +1 / 4 -1 / 4 0 (The contribution here is 1 / 4 for each.)

[0085] This is a correction applied to the input and the result of the transformation (estimation procedure). First, the result is calculated as follows:

[0086] The input values ​​for yellow and magenta are corrected from 50%+50% to 49.5%+49%.

[0087] Subspace Table 1: Interpolation for printing yellow 49.5% Yellow 92.5050 (Very bright, not reddish, moderately yellowish)

[0088] Subspace Table 2: Interpolation of Magenta Ink 49% Magenta 75350 (Slightly bright, moderately reddish, and not yellowish)

[0089] The estimation method is provided by combination (from the XYZ color space). 49.5% Yellow + 49% Magenta 733441

[0090] Final result after target value correction: Yellow 50% + Magenta 50% 73.2533.7541 (Interpolation triangle added)

[0091] This invention describes a practical and feasible solution for those skilled in the art that enables sufficiently fast and high-quality color space conversion to a color space, even with very high-resolution layouts and a large number of input colors, in a reliable printing system. [Brief explanation of the drawing]

[0092] Further advantages and features will be explained below with reference to the drawings. The drawings are as follows: [Figure 1] Corrections on both the supplier and recipient sides for conversions based on the recipient's values. [Figure 2] Correction nesting until the desired accuracy is achieved. [Modes for carrying out the invention]

[0093] Figure 1 shows the steps of a correction method for a selected color dataset Q1 from n colors in the source color space (101) having color components q(1), q(2), ..., q(n), involving iterative adjustment of the source data followed by residual correction of the destination data. For this color dataset Q1, a verified and stored destination value dataset Z1 of m values ​​in the destination color space (102) having components z(1), z(2), ..., z(m) is made available by measurement. In step 103, the current source color dataset Q2 is initialized with the starting value (101). This is transformed using the transformation rule, e.g., model, provided in step 104. A dataset Z2 of m values ​​in the destination color space 105 having components z'(1), z'(2), ..., z'(m) is generated.

[0094] In step 106, the differences between individual values ​​are calculated from Z1 and Z2, respectively. A check is performed in step 107 to confirm that these differences are sufficiently small. If not, the color dataset Q1 from step 103 is corrected in step 108 to generate color dataset Q2. The corrected color dataset is transformed in step 104, and steps 105-107 are repeated. The steps are shown schematically, and the actual procedure concerns some of the optimization methods used, particularly how the correction is determined in step 108. Finally, if the difference is minimal, the difference between source color datasets Q2 and Q1 is calculated in step 109. For example, a transformation table of numerous pairs of Q1 and Z1, the difference between Z1 and Z2, and the difference between Q2 and Q1 can be maintained and interpolated. The source difference is applied to the input data in the overall transformation 110, followed by the application of the original transformation 104 to the destination color space and the final correction of the residual error 106.

[0095] Figure 2 refers to Example 2 in the explanation starting on page 15. It describes multiple hierarchically refined corrections of Model TRV201 using measurement data from a printed CMYK test chart 202. For the initial correction at the corners of the CMYK color space, 2% and 100% are used. 416 combinations and their measurement data are acquired in process 203. For each combination (204), the supplier and recipient corrections are determined (205) as shown in Figure 1, and are compiled into tables, resulting in total tables A1 (206) and B1 (207) with 16 entries each. These are combined with the model transformation rule TRV to form a new corner correction transformation TRV1 (208).

[0096] The second correction uses the 0, 40, and 100% CMYK combinations available in many test charts (209) and their measurements. 3 4 For each of the 81 combinations (210), the correction for TRV1 is determined (211) and stored in tables A2 (212) and B2 (213). Since the corners are already correctly corrected in TRV1, these 16 of the 81 correction entries are zero. A2 and B2 are positioned as the next correction shells around TRV1, and there are two shells around TRV (214).

[0097] The TRV2 model (214) corrected in this way is accurate in all corners and midtones, requiring only minor adjustments. Furthermore, the shell uses combinations of 0, 20, 40, 70, 100 for CMY and 0, 20, 40, 60, 100 for K in the case of ECI2002 or IT8.7 / 4 test charts, and 0, 10, 20, 30, 40, 55, 70, 85, 100 for CMY and 0, 10, 20, 40, 60, 80, 100 for K. This makes optimal use of the grid structure of these standard test charts. Once the model is corrected in this way, the CMYK print data of the project can be spectrally simulated and displayed on a screen or inkjet test printing system using standard procedures.

Claims

1. A method for performing computer-assisted conversion of color data from a source color space to a destination color space using a given conversion rule TRV, The color space of the supplier includes n colors that exist in combinations of color parts q(1) to q(n) in each dot to be printed. The color space of the supplier includes m values ​​that can be combined to form a combination of components z(1) to z(m), At least some of the color component combinations q(1), q(2), ..., q(n) of the supplier's color space are assigned to combinations of components z(1), z(2), ..., z(m) of the destination's color space. The aforementioned method, a) A step of selecting a color dataset Q1 of n colors from the supplier's color space having color components q(1), q(2), ..., q(n), to which a value dataset Z1 having combinations of components z(1), z(2), ..., z(m) of the supplier's color space is assigned. b) A step of generating a color dataset Q2 of n colors from the source color space, which is initially set to be the same as Q1 and has color components q'(1), q'(2), ..., q'(n), c) A step of converting the color dataset Q2 using the conversion rule TRV to generate an actual value dataset Z2 of m values ​​of z'(1), z'(2), ..., z'(m) in the color space of the supply destination, d) A step of modifying the color dataset Q2 based on an n-dimensional optimization procedure, wherein the n-dimensional optimization procedure is applied by repeating step c) to minimize the difference between the nominal value dataset Z1 and the actual value dataset Z2 in least squares error, so that Q2 ultimately includes the optimized color components q'(1), q'(2), ..., q'(n) obtained as a result, e) A step of calculating the difference Dq(i) between each color portion of the n colors in the optimized color dataset Q2 and each color portion of the color dataset Q1, using the formulas Dq(1) = q'(1) - q(1), Dq(2) = q'(2) - q(2), ..., Dq(n) = q'(n) - q(n), f) The process of calculating the difference Dz(j) between each component of value dataset Z1 and each component of value dataset Z2 using the formula Dz(1) = z(1) - z'(1), Dz(2) = z(2) - z'(2), ..., Dz(m) = z(m) - z'(m), g) A method characterized by including the step of applying the difference Dq(1), Dq(2), ..., Dq(n) and the difference Dz(1), Dz(2), ..., Dz(m) when converting a set of n color data from the supplier's color space to m values ​​in the destination's color space, by adding the difference Dq(i) to the n color data q(i) of the supplier's color space for each value from i=1 to n during the conversion of a dataset. The color data q'(i) thus converted to m values ​​z'(j) for values ​​from j=1 to m in the destination's color space according to the conversion rule TRV, and each component z'(j) of the corresponding difference Dz(j) is added to each component of z'(j).

2. The method according to claim 1, characterized in that when determining the setpoint dataset Z1, the dataset Q1 is printed out and the setpoint dataset Z1 is determined by spectral measurement.

3. The method according to any one of claims 1 or 2, characterized in that in step d), only the color components of n colors that are not zero change.

4. The method according to claim 1 includes a plurality of sets of color data Q ｊ The method is performed on a plurality of sets Q ｊ The difference Dq(i) determined for ｊ and Dz(i) ｊ The method according to any one of claims 1 or 2, characterized in that the differences are managed in a conversion table, and in the conversion of step f), the differences to be applied are interpolated in these conversion tables.

5. The method according to any one of claims 1 or 2, characterized in that the method is performed on a computer unit by control software, the computer unit comprises an input unit that provides digital color data of the source color space of the project, an output unit that outputs converted values ​​of the destination color space, and a memory that stores a conversion table, and the values ​​of the destination color space are generated by the control software for inputting the project's color data using the conversion table and provided as a dataset.

Images