Abstract
At VSS 2005 we described how to use the Guided Search model of visual search to optimize colors on a simple mall directory. We now show how to apply a similar idea with a more general model and more complicated stimuli. The stimuli are similar to maps of parks, with various items of interest indicated by colored icons. We used the model to identify the optimal selection of colors and icon shapes that would promote fast search of the maps. The model analyzes the color, shape, and eccentricity of a target item relative to other display items on the map image and weights computational measures of these effects to predict search time. The model weights were parameterized to fit reaction time data from an experimental study where observers searched for a target in a map with different sets of icons. 45 observers in 3 groups each viewed 150 maps. The model did a good job (r=.67) fitting the average reaction times of the 450 total trials. We then used the model to explore optimization tasks that varied the color palette that could be used, the distribution of icon search frequencies, and the shapes of the icons. We also explored the generalizability of the model's parameterization by applying the model to a different map design task. This approach shows good promise for applying theories from visual perception to design tasks. The study highlights deficiencies in many current models of visual search that cannot be applied to this kind of project.