The statistical saliency model can choose colors for items on map displays

When a display is cluttered, drawing attention to specific display items may be difficult. Our current project examines how to choose distinctive (i.e., salient) colors for items on high-clutter and low-clutter maps using predictions of a model of visual search. Thirty observers rated the degree of clutter on each of 150 MapQuest maps containing symbols representing colored location pushpins. We calculated the average clutter rating for each map and identified 20 high-clutter maps and 20 low-clutter maps. Next, the Statistical Saliency Model (Rosenholtz et al., 2007) was used to choose colors for pushpins on each of the 20 high-clutter and 20 low-clutter maps. For each map, we calculated the model-predicted salience of a pushpin added to the map, given different potential pushpin colors (the 267 colors in the ISCC-NBS standard color system). We chose the color with maximum predicted salience. For comparison, we also created maps where the pushpins were assigned the color with minimum predicted salience or median predicted salience. We followed this procedure for all 40 maps, resulting in a set of 120 maps. We validated the color assignments in a visual search experiment where observers were shown a pushpin and then asked to search for it on a map. Maximum predicted salience pushpins were found faster than pushpins of other colors on both high- and low-clutter maps, indicating that our approach can choose colors for display items to facilitate search. Meeting abstract presented at VSS 2013