An intuitive color-selection tool

A color-picker or color-selection tool is part of a GUI that allows users to select colors for use in software applications. There is a widespread belief that some color spaces or models are more “natural” than others. However, this intuitively appealing idea that a color space based on the nature of color perception (such as HSL) may be preferable to one that is driven by the nature of the technology (such as RGB) has not been confirmed by all studies. In this paper we argue that it is not the color space per se that is the most important factor underlying the usability of a colorselection tool but rather the color-mixing algorithm. An experiment was conducted to determine matching performance in a color-selection tool where the sliders interacted with the on-screen color using either a direct additive model or an indirect subtractive model to control the RGB values of the on-screen color. When observers were given a limited amount of time to use the sliders to match target colors their performance was statistically superior when they used the subtractive CMY sliders than when they used the additive RGB sliders. This is consistent with some previous work that suggests that observers possess better internal models of subtractive mixing than additive mixing and that the design of color-selection tools could exploit this. It should be noted that this work differs from some related work that has looked at the influence of color space on the usability of colorselection tools. Our hypothesis is that it is the color-mixing model that relates the slider bars to the on-screen color that is important rather than the choice of color space itself. Introduction A color-picker or color-selection tool is part of a GUI that allows users to select colors for use in software applications. A number of studies have considered whether the choice of color space affects the usability of such tools. There is a widespread belief that some color spaces or models are more “natural” than others. For example, one established text claims that “The RGB, CMY, and YIQ models are hardware oriented. By contrast, ... HSV ... is user-oriented, being based on the intuitive appeal of the artist’s tint, shade and tone”. However, this intuitively appealing idea that a color space based on the nature of color perception (such as HSL) may be preferable to one that is driven by the nature of the technology (such as RGB) has not been confirmed by all studies. For example, Douglas and Kirkpatrick concluded that choice of color space (specifically, in this case, comparing RGB and HSL) was not an important factor in the usability of a colorselection interface. In this paper we argue that it is not the color space per se that is the most important factor underlying the usability of a color-selection tool but rather the color-mixing algorithm. The use of RGB sliders as part of a color-selection tool is essentially a color-matching task (where the user is additively combining primaries to match a target color) and therefore the user must have knowledge about the additive-mixing properties of the primaries controlled by the sliders. In a previous study we showed that users are better able to predict the results of subtractive color mixing than they are additive color mixing. Users may possess well-developed internal models of subtractive color-mixing processes that developed in childhood as they experimented with inks and paints. A typical user, for example, would not be surprised to be informed that yellow and blue inks mixed together make green but may not be so familiar with the fact that red and green lights can be added together to make yellow. In this work, user performance in color-selection tasks is measured and compared for color-selection tools that are based upon additive and subtractive color mixing. Method A graphical-user interface (GUI) was created in MATLAB that enabled observers to adjust a color by adjusting each of three slider bars (see Figure 1). Users were instructed to adjust a sample color so that it was a visual match for a target color and were given a fixed amount of time to complete the task. Two time limits were employed: 120 seconds and 30 seconds. A total of nineteen observers were recruited to take part in the experiments but three were discarded because they displayed abnormal color vision. Six observers (3 male, 3 female) with normal color vision took part in the 120-second experiment. A different set of 10 observers (5 male, 5 female) with normal color vision took part in the 30second experiment. Figure 1: MATLAB GUI for color selection. The left-hand color is adjusted by the sliders to match the right-hand one. The bar at the bottom indicates the time remaining for the task. In the experiments, a total of 12 target colors were presented in turn in the right-hand part of the display (see Figure 1). The RGB values of the target colors are given in Table 1. The first three of these were for training purposes only to allow users to get used to the slider bars. The accuracy of the matching task for the first three colors in Table 1 was therefore not considered and results were calculated for the remaining nine target colors. Table 1: sRGB specifications of color targets used in the