D. Novaković, N. Milić: Effects Of Display Gamma..., acta graphica 20(2009)1-4, 21-29 udc 655.3.062.22 original scientific paper received: 31-05-2009 acta graphica 176 accepted: 10-06-2009 Effects Of Display Gamma And Illuminant On The Appearance Of Colours Viewed On Lcd Monitors Author Dragoljub Novaković*, Neda Milić The Faculty of Technical Science, Departman for Graphic Engeeniring and Design Novi Sad, Serbia *E-mail: [email protected] Abstract The ciecam02 colour appearance model was developed to allow prediction of accurate colour appearance under various viewing conditions. In this study, a soft- ware module, based on mathematical transformations of the ciecam02 model, was designed and implemented. Unlike the previous software modules, which were allowing use of only a limited choice of common standardized icc display profiles, the developed module also allows the use of optimized profiles created by the user. This paper presents prediction of the appearance of test images with assigned dif- ferent display profiles after changing illuminant. The two used optimized profiles, created by software characterization of specificlcd display, have a different display gamma value (1.8 and 2.2). It is observed that, after changing the initial cie d50 to a higher colour temperature illuminant (d65), the colours of images will be shifted to the blue hues, and, in the case of illuminant temperature less than d50 (a and f11), to yellow hues. It is also observed that the intensity of colour shifts in the simulated images with associated profiles which have different gamma value will be significantly different, which implies that theicc display profile and display gamma value have the great impact on the prediction of the accurate colour ap- pearance under various viewing conditions. Keywords: ICC profile, gamma, ciecam02 colour appearance model, viewing conditions 21 D. Novaković, N. Milić: Effects Of Display Gamma..., acta graphica 20(2009)1-4, 21-29 1. Introduction ance of maximum performance achievement of device with respect to the given parameters. The colour appearance models represent The optimizedicc display profiles are used when mathematical models which, through a series of we want to completely accurately characterize a transformations (a linear transformation from specific monitor, which is of great importance xyz tristimulus values to cone responses, the for a consistent soft proof (Sachs 2008). chromatic adaptation transformation, transfor- mation to opponent and achromatic responses Monitor calibration and profiling ensure that etc.) calculate the colour values in order to the white point of display, rgb primaries and predict the colour appearance across disparate gamma value are set correctly (Sachs 2008). The viewing conditions, after the adaptation of the value of gamma, as one of the basic parameters human visual system to changes. In graphic that are defined during calibration of the moni- industry is currently used and widely accepted tor, has a great impact on the appearance of CIECAM02 model, defined by CIE organization the colours. It describes non-linearity between (2002). Using ciecam02, the following percep- the digital signal and the luminance values of tual colour attributes can be calculated: the ab- colours on the display (the tone reproduction solute color appearance attributes of brightness curve of signal). In case of the gog model of (q), colorfulness (m) and hue (h) and the rela- calibration of crt monitors, this non-linearity tive color appearance attributes of lightness (j), is defined as (cie technical report 122-1996): chroma (c), saturation (s), and, again, hue (h). Then, with the inverse transformation from (1) perceptual attributes xyz values are com- puted, taking into account the given fi- where l represents the normalized display lu- nal viewing conditions (Fairchild, 2005). minance, v is the normalized input signal of red, The cie Lab colour space can be considered a green and blue channels, γ is the value which colour appearance model because it includes describes the non-linearity in the relationship simple chromatic adaptation transforms and pre- (gamma), a and b are the values of gain and dictors of lightness, chroma and hue. However, offset of the input signal, respectively. Unlike the Lab colour space does not have the capabil- crt monitors, which were originally non-linear ity to include the effects of the background and display devices, the lcd displays have natural surround in which the stimulus is observed, so linearity in relationship between the input sig- it is not sufficient to predict the impact of view- nal and the display luminance. However, manu- ing conditions. Therefore, there is a tendency to facturers of lcd monitors intentionally build in implement ciecam02 model in the workflow circuitry that changes the display’s natural lin- of colour management systems (Fairchild 2005). ear light output characteristic into a non-linear In colour management system, profiles are used characteristic in order to mimic the gamma of to describe the condition of some device and crt displays (Schulte 2004). In the case of the its capabilities in terms of colour reproduction. lcd monitors, equation (1) is, due to produc- For graphic applications, profiles recommended tion technology, simplified: by cie, fogra and similar institutions or spe- cially generated for a particular device by the manufacturer, are often used. The advantage of (2) using these profiles is reflected in the possibil- ity of their use for the class of the device and Gamma value defined by equation (2) is the the specific work requirements. However, these physical (original) gamma of monitors and for profiles do not show the actual reproduction most monitors this value is 2.5. Adjusting lut of capabilities of device. Using optimized profiles graphic card original gamma can be customized in colour management systems (profiles gener- to the desired value. Corrected gamma value is ated intentionally for specific device) instead of usually called a display gamma. In Windows op- standardized ones, involves a number of advan- erating systems, the display gamma of 2.2 is rec- tages, with the main advantage being the assur- ommended, whereas in case of the Mac systems, 22 D. Novaković, N. Milić: Effects Of Display Gamma..., acta graphica 20(2009)1-4, 21-29 this value is slightly lower and amounts to 1.8. In 2. Method accordance with (2), the value of gamma has the largest influence on the reproduction of middle Before the beginning of testing, software tones, while the colours are generally perceived calibration and characterization of specific lcd as darker with an increase in the gamma value monitors used in this study (Samsung t220) was (Poynton 1996). carried out. With software calibration (d50, 100 2 Soltic and others (2004), based on math- cd/m ), display gamma value is set to 1.8 in the ematical calculations, showed that the change in case of the firsticc display profile, and to 2.2 in the gamma value leads to changes in the percep- the case of another icc profile. tual attributes of colour defined by ciecam02 The values of gamma 1.8 and 2.2 were selected model, using the matrix conversion from rgb on the basis of the recommended values for to xyz colour space of standardized srgb pro- Windows and Mac operating systems, respec- file. tively. Bodrogi et al. (2002) concluded that there In the case of both icc profiles, matrix con- are significant differences between the colours version from one colour space to another is when using standardized srgb tone reproduc- used. In other words, one matrix describes the tion curves of channels instead of actually meas- conversion of each colour that the monitor is ured curves. He and others (2008) have pointed able to display. Conversion matrix is represent- to significant colour differences defined in the ed in equation (3): cie Lab and ciecam02 colour spaces as a result of changing the system gamma of lcd display. (3) Besides the viewing conditions, input param- eters of ciecam02 model are xyz tristimulus values of image pixels, which are determined by where: icc profile assigned to image, whereas the pre- vious software were allowing use of only a limit- X, Y, Z – tristimulus pixel values ed choice of common standardized icc display Xr, Yr, Zr - maximum tristimulus values r chan- profiles. A software module was developed and nel implemented, which, in addition to standard, Xg, Yg, Zg - maximum tristimulus values g allows the use of optimized display profiles cre- channel ated by the user. Xb, Yb, Zb - maximum tristimulus values b channel In this study is simulated the appearance of r , g , b - digital 8-bit values for red, green images with assigned optimized icc display 8bit 8bit 8bit and blue channel profiles which have different display gamma, af- γ - gamma. ter changing the initial cie d50 illuminant to d65, f and a illuminants, based on the ciecam- Matrix elements of created icc profiles are 02model. It was assumed that the increase in il- given in: luminant temperature will lead to colour shifts to the blue hues and the decrease will cause the (4) shifts of colour to yellow hues. It was also as- sumed that there will be the significant colour differences between the simulated images with Characterization of the monitor, besides the associated profiles which have different gamma conversion matrix, includes tone reproduc- value. The rgb values of processed images are tion curves- gamma function of each channel converted to the Lab values and the position (Burns, 2002). of the colours in the images was observed and compared in the cie Lab colour space. Meth- After that, profiles are assigned to test images ods, as well as presentation and discussion of and images are processed using a software mod- results, are defined below. ule based on the ciecam02 model in order to 23 D.