Diogo Maurício Gonçalves

Digital Color Conversion with ICC Profiles in Skin Tones


Fernando Carvalho RodriguesIADE – Instituto de Arte Design e Empresa
Maria Cristina Ventura, ISEC – Instituto Superior de Educação e Ciências


Evaluate and compare the results of color conversions with ICC profiles in a color sample representative of the human skin tones. The color deviations will be calculated using the DeltaE formula CIEDE2000. The results will be systematized and treated quantitatively, focusing on (i) calculating a set of quantitative indicators, (ii) performing a set of statistical tests in order to obtain validity and confidence in the results obtained, and (iii) using artificial neural networks – ANNs – to find patterns and predictable behaviors in color conversion methods.

Keywords: Color Management, Color Profiles, CIEdE2000, ANNs, Skin Tones

1. Introduction

Nowadays, DCM software is ubiquitous in all digital systems in order to enable automatic color conversions and color representations. In a DCM system (Digital Color Management), color conversions are enabled by small digital files, the Color Profiles. They are similar to a color catalog, encompassing the information about the colors that are possible to reproduce or to achieve in a specific device or color space. The DCM controls the Color Profiles with a set of commands, the tuning of the color transformation. In these commands there are at least four different algorithms to calculate the best colors and also an important option, the BPC (Black Point Compensation), used to recalculate the darkest possible colors. In a DCM system, the combination of algorithms and BPC can produce a minimum of six different results for each Color Profile.

2. Trial Experience

We measured and evaluated the performance of these semi-automatic DCM color conversions using a sample of colors similar to the variety of human skin tones, considering the qualitative assessment the human sensibility to color differences.

We evaluated color conversions from a color sample in RGB (Red, Green and Blue) to 12 CMYK (Cyan, Magenta, Yellow and Black) Color Profiles commonly used in the printing industry. In each conversion we compared the six different algorithm combinations, a total of 72 results. The color sample is intended to be representative of the diversity of colors of the humankind, with 156 colors from pale to dark tones, mostly from a color palette of the brand “Pantone Skintone”TM.

The color deviations were calculated with the DeltaE “CIEDE2000” formula recommended for the printing industry. The results were systematized and treated quantitatively with major focus on (i) calculate a set of quantitative indicators and (ii) perform a set of statistical tests in order to impart meaning and guarantee the reliability statistical conclusions obtained, that included: (i) Test Pearson Correlation Coefficient to evaluate linear relations between batches; (ii) Test Standard Deviation and Coefficient of Variation to evaluate data dispersion; (iii) Test Null Hypothesis Wilcoxon to evaluate the independence of groups of data; (iv) Test Null Hypothesis ANOVA to estimate the representability of closely mean values.

3. Results of Trial experience

Our data shows that considering the human sensibility to color difference, the results can be quite different when choosing different algorithms. Overall, in the six possible algorithm tunings evaluated, (i) none of the algorithms was always the best, (ii) two didn’t achieve any good result and (iii) the best algorithm combination only achieved good results in half of the Color Profiles.

4. Conclusion of Trial Experience

Restricted to the universe of human skin colors and Color Profiles tested, we think that the mindset in the graphic art industry where the choice of a good Color Profile is enough to ensure quality needs some rethinking, because the algorithm choices that fine tune the color conversions seems to have a major and unpredictable influence in the final result. These results and respective analysis where of a major importance to our Phd Thesis in order to (i) validate the methodology approach to the experience and, (ii) using the results to choose the best approach and choice of the Artificial Neural Networks algorithms to use. At the moment, we are presenting the results of the Trial Experience in conferences in order to validate our approach and collect inputs.

  1. Berns, R., 2000. Principles of Color Technology - Billmeyer and Saltzman’s 3rd ed., New York: Wiley.
  2. Chen, S. & Beghdadi, A., 2011. Improve gamut mapping of Color Management System by perceptual quality-oriented analysis. In 3rd European Workshop on Visual Information Processing. IEEE, pp. 12–17. Available at: [Accessed May 29, 2016].
  3. Fairchild, M., 2013. Color appearance models 3rd ed., Chichester West Sussex U.K.: John Wiley & Sons.
  4. Foster, D.H., Nascimento, S.M.C. & Amano, K., 2005. Information limits on identification of natural surfaces by apparent colour. Perception, 34(8), pp.1003–1008.
  5. Green, P., 2010. Color management: understanding and using ICC profiles, Chichester West Sussex U.K.: John Wiley & Sons.
  6. Homann, J.-P., 2010. Digital color management principles and strategies for the standardized print production, Berlin, Germany: Springer.
  7. Hunt, R. & Pointer, M., 2011. Measuring colour 4th ed., Chichester West Sussex U.K.: Wiley.
  8. Saguer, M.M., 2013. Color and Imaging Conference: ingentaconnect Table Of Contents. In Validating the black point compensation standardization. pp. 114–117. Available at: [Accessed September 28, 2015].
  9. Sharma, G., 2005. The CIEDE2000 color-difference formula: Implementation notes, supplementary test data, and mathematical observations. Color Research & Application, 30(1), pp.21–30. Available at: [Accessed July 15, 2015].
  10. Tastl, I. et al., 2005. ICC Color Management and CIECAM02. In Color and Imaging Conference. Society for Imaging Science and Technology., pp. 217–223. Available at: [Accessed July 31, 2015].