Experimental Investigation of Distortion of Colour Harmony: A Harmony Distortion Index

This study investigates the distortion of visual colour harmony impression in case of different types of artificial changes among the perceptual attributes of test samples. A set of visual experiments were carried out to investigate the observer’s colour harmony impression as a function of artificial distortions. The visual colour harmony impression of a colour combination appearing harmonious under a reference illuminant might decrease if the combination of samples is illuminated by different test light sources, especially RGB LED clusters. A new quantity, the so called Harmony Distortion Index (HDI) is defined to describe the harmony distortion property of light sources. Introduction Authors investigated existing colour harmony models[1,2,3] and developed several new predictive formulae[4] to quantify visual colour harmony impression. A Harmony Rendering Index (HRI)[4] based on the harmony predictions of colour harmony formulae for a set of test colour combinations under different illuminants was also developed, to describe light source colour quality. In an earlier experimental study[5], authors came to the conclusion that colour harmony impression decreased significantly by changing the reference light source to a white LED lamp consisting of a cluster of RGB LEDs of unusual spectral power distribution. A possible explanation of this phenomenon was claimed to be that there were very different and non-systematic colour shifts, concerning both magnitudes and directions among the samples[5]. In a previous study[6], several “equivalent” shifts were displayed to 7 observers in order to investigate the authors’ “shift-invariance” hypothesis of colour harmony. In case of “equivalent” shifts, each of the three perceived attributes (hue, chroma and lightness) changes to the same extent for all colours of the colour combination. Several “non-equivalent shifts” were also shown to examine the distortion of perceived colour harmony. In this article, these experimental results are analyzed to form a model describing the colour harmony distortion property of light sources. Experimental Method To investigate the distortion of visual colour harmony, a series of visual experiments were carried out. Equivalent and non-equivalent distortions were simulated on a calibrated CRT monitor changing hue, chroma and lightness separately (see Figure 1). For “equivalent” shifts, each of the 3 perceived attributes (CIECAM02[7] hue, chroma and lightness) changed to the same extent for all constituent colours of the colour combination. “Non-equivalent shift” means that the 3 perceived attributes change not to the same extent or not to the same direction for the two or three constituent colours of the colour combination. The extent of distortions we applied were 2, 4, 7, 9, 12 CIECAM02 ΔE units. The task of the 7 observers (3 male, 4 female, aged between 22-24) was to evaluate visual colour harmony on a scale from -5 to +5, where -5 means very disharmonious, +5 excellent harmony. Observations were repeated three times, to get significant experimental results. Before the experiments, the observer’s colour vision was tested by the Farnsworth-Munsell 100 Hue test. Figure 1.: Experimental images on the CRT monitor in case of two and three colour combinations Results From the analysis of visual results, the relation between distortions and perceived colour harmony impression was determined. As can be seen in Figure 2 distortion of visual colour harmony is different in case of artificial equivalent and nonequivalent distortions. On this Figure, the effect of the size of the artificial distortion can be also seen. In case of “small” distortions, the extent of the artificial distortion is 7 CIECAM02 ΔE units. For “large” distortion, this distortion is 12 CIECAM02 ΔE units. The main factors which are responsible for the distortion of colour harmony impression were identified as: 1. extent of distortion 2. perceptual attribute 3. a further quantity describing the “equivalent/nonequivalent ratio” of the distortion. Figure 2.: Distortion of observer’s visual colour harmony in case of different type of artificial distortions for two colour combinations. As a conclusion of these first results it can be stated that the distortion of visual colour harmony can be predicted with the help of a new metric, which is composed of functions which are good descriptors of the effect of hue, chroma and lightness distortions separately. Mathematical functions were fitted to experimental data to describe the effect of these factors. These functions must be also responsible for predicting the distortion of visual colour harmony as a function of the extent of distortion. Visual results and fitted curves can be seen in Figures 3-5.