Building Better ICC Profiles with X10 Media Manager

Transcription

1 ONYX White Paper Building Better ICC Profiles with X10 Media Manager January 2011 ICC profile generation is an important part of developing and using a color managed workflow for printing. With an ICC profile, the characteristics of how a print mode reproduces color are captured and leveraged in order to provide predictable reproductions of images. In this white paper we will take a closer look at ICC profile generation in version X10 of Media Manager. We will discuss several changes in how ICC profiles are built in order to develop an in-depth understanding of the profile generation approach utilized in Media Manager. Basics of ICC Profile Generation The following basic steps are used to build an ICC profile for a printer. These steps are generally followed by anyone who generates printer ICC profiles: 1. Various color combinations are printed. 2. A model is created that can be used to convert device channel output combinations to expected color output. 3. The model from step 2 is used to populate the device values to PCS (profile connection space) tables. 4. An Ink separation approach is put into place (black generation and HiFi color usage). 5. The model from step 2 is used to determine the gamut available to the device, and a gamut mapping strategy is put into place. 6. The model from step 2, the ink separation from step 4, and the gamut mapping strategy from step 5 are used to build the PCS to device value tables. 7. The tables created in steps 3 and 6 as well as various metadata are saved as an ICC profile. The differences between ICC profiles that are generated by different profile building packages primarily relate to the choices made in steps 2, 4, 5, and 6. Device Modeling One of the primary motivating factors behind developing our own color profiling engine was to overcome limitations we found in existing profiling packages. Specifically, when trying to maximize CMYK color space output gamuts, the use of selective ink limiting compromised the models assumed by the profiling package being incorporated at the time. The following simple example may help. 1

2 Imagine that you are trying to model the density of an ink. As you have more ink you generally get greater density. A model is generally used to predict the expected density for any amount of ink desired. If color management decisions are made based upon the model, then your color management will only be as good as the model. The following plot shows a hypothetical device behavior. To create a density model, you print and measure several patches of ink, and then use that information to create a model. There are several approaches that can be taken. One approach is to describe things using a mathematical model. In this example, one could use a math function like y=ax b +c. The measurements are used to estimate the constants (a, b, and c) in the function. If the math corresponds well to how the device behaves, then not very many measurements are needed. However, the model s ability to predict the actual density may suffer if the math model does not correspond well to the behavior of the device. In this case, adding more sampling points will not improve the results. To get an improvement you need a different math model. Another approach is to assume a linear relationship (i.e. local linearity) exists between measurements. Errors in estimation occur in places where the device is not locally linear. Errors can be reduced by increasing the sampling of the device behavior. In the next case the number of patches is 2

3 doubled, and except for the very darkest regions there is nearly a perfect fit of the device behavior. Note: An additional sampling step in the darker regions would have resulted in an excellent fit. Based on the number of patches available and other print mode properties, like the use of gamut expanding Spot or HiFi colors (i.e. Orange, Green, Red, Blue), the ONYX ICC profile build engine dynamically selects an interpolation model or a functional model. Measurement Error One difference between a functional model and an interpolation model is how incorrect measurements affect them. In the following example, one of the measurements is incorrect (for some reason). In a functional model, the predictions for all ink percentages are affected. In the interpolation model, all the measurements are assumed to be correct. Therefore, only the ink percentages near the incorrect measurement are affected. In looking at these results it is also important to note that the predictions of the functional model will be smooth. However, the accuracy may not be so good. Conversely, the smoothness of the interpolation model is to a greater extent determined by the smoothness of the device behavior combined with the smoothness of the measurement data. For this reason it might be a good idea to measure samples multiple times (with averaging) to reduce the affects of variability in measurements. Note: Variability in measurements is most likely to occur at lower resolutions or in cases where the measured surface is not very smooth. When measuring an ICC swatch, it is important to make sure that measurements are performed correctly. In some cases incorrect measurements can be caused by measuring the wrong patches or problems with the measurement device recognizing patches in a strip. In X10 a new feature was added that highlights measurements that are significantly different than normal. The following case shows an example where a patch was incorrectly measured twice. 3

4 After measuring an ICC swatch, it is important to review the measurements to determine if there are any significant differences between the expected and measured values, and re-measure patches/strips that have been measured incorrectly. Note: Highlighting is only an expression of a potential problem. One should look for shifted color patches or groups with obviously incorrect hue that may indicate there is a problem. In other cases the output device may simply have a color gamut that is somewhat different than normal. In such cases the profile measurement data will correctly account for this behavior. Gamut Mapping Before gamut mapping can be performed, a gamut boundary must be determined. This is done by taking the set of all device colors that are possible with the ink separations defined in the ICC profile settings. Both black generation settings as well as HiFi color separations are taken into account by the ink separation algorithms. Different settings will change the effective gamut boundary. With a gamut boundary determined, the colors inside the gamut are established by using the printer module set up using the measured patches. Colors outside the gamut boundary cannot be reproduced by the print mode. For each color outside the print mode gamut it is necessary to choose a wrong color to use instead. Gamut mapping is the process of deciding what wrong colors to use. There are many ways to do this leading to different gamut mapping selections in the ICC profile build options. The X10 profile build options now include Enhanced Color gamut mapping strategy. This gamut mapping strategy makes assumptions about colors that are inside the gamut of an RGB color space but are outside the print mode s gamut. These colors are particularly mapped to enhance the chroma achieved by allowing for subtle changes in hue. This results in a more chromatic (and possibly more pleasing) output when printing RGB images using v2 colorimetric based RGB profiles (most notably the srgb_iec61966_21.icm and AdobeRGB1998.icc profiles). However, v4 RGB input profiles may already perform this operation and little difference will be noticed since the input profile will be performing the color rendering (Example: the srgb_v4_icc_preferrence.icc profile). When analyzing differences in gamut mapping strategies, it is important to recognize that no single gamut mapping strategy will be best for all circumstances. The advantages that one gamut mapping strategy has over other strategies can often be image dependent. What may look good for one image may be inappropriate for another. In other cases differences between strategies may not be apparent depending upon the image involved. Therefore, 4

5 differences when using the Enhanced Color gamut mapping mode as opposed to the other strategies can be subtle at times. Differences really need to be evaluated on a per image basis with actual printed output. This gamut mapping strategy has no actual information about the input gamut that will be connected to it. When using CMYK input profiles with an output profile built using the Enhanced Color gamut mapping strategy that has a smaller print gamut, some hue shifts may be noticed. Making use of this new gamut mapping strategy requires rebuilding the output profiles. This does not require reprinting or re-measuring an ICC swatch. Simply edit the ICC profile in X10 Media Manager and change the gamut mapping setting to Enhanced Color, and rebuild the ICC profile. It may be a good idea to make a copy of the Print Mode before doing this. Conclusions The accuracy of an ICC is only as good as the model used to predict the behavior of a print mode. Both the input and output tables are based upon this model. As such it is possible to have a profile that internally has excellent predicted behavior yet is significantly inaccurate. Both functional and interpolation device modeling approaches have been discussed. To get the best results for CMYK process modes, it is recommended that the swatch with the most patches be used as this will best adapt to varied print behavior. Care should also be exercised to ensure that measurement data is correct. It is strongly recommended to check the read swatch dialog for anomalies before building profiles, and it is also suggested that multiple readings be made in conditions where great variability in measurements are possible. Using the Enhanced Color gamut mapping strategy can result in improved output when printing RGB images. ONYX Graphics Inc S. High Tech Drive Salt Lake City, UT Phone: Toll Free: ONYX, the ONYX logo, ONYX PosterColor and ONYX PosterShop are registered trademarks, and ONYX ProductionHouse, ONYX RIPCenter and ONYX GamaPrint are trademarks of ONYX Graphics, Inc. All product and company names are trademarks or registered trademarks and are the property of their owners and are respectfully acknowledged. Copyright 2011, ONYX Graphics, Inc. All rights reserved. 5