Influence of local scene colour on target detection tested by global rearrangement of natural scenes

Local scene colour can influence the visual detectability of an object or target, but so can the familiarity, meaning, and global organisation of the scene. The aim of this study was to test whether the effects of local scene colour on target detectability are secondary to global effects. A target-detection task was undertaken by human observers with coloured images of natural scenes that were cut into quarters, randomly rearranged, and then reassembled. The target was a small, shaded, neutral grey sphere located randomly within the scene and matched in mean luminance to its local surround. It was found that observers’ target-detection performance with the rearranged images was about as good as with the original images. The combination of local colour properties, namely, lightness and the red-green and yellow-blue components of chroma, accounted, respectively, for 55% and 50% of observers’ detection performance with the original and rearranged images. Despite the disruption of global organisation, local scene colour continued to influence target detection. Introduction Although an everyday task, visually detecting an object or target in a scene presents a challenge for analysis. This is because performance depends on the structure of the scene being searched as well as on its contents [1-3], both of which can vary markedly in the natural world. It is often assumed that spatial achromatic image features of scenes, such as edges, texture, orientation, and contrast [4-6] are the main determinants of target detectability, but local scene colour can also be influential, accounting for about 60% of the variance in observers’ performance with natural scenes [7]. Yet it is unclear whether the effects of local colour properties in natural scenes are truly local, in the sense that detection still takes place within a coherently structured environment, with all the advantages of familiarity, meaning, and global scene organisation that are immediately apparent to an observer [810]. It is possible, therefore, that the effects of local colour properties on target detection are secondary to the effects of the more global properties of natural scenes. The aim of the present study was to test this alternative interpretation. A target-detection task was undertaken by human observers with coloured images of natural scenes that were cut into quarters, randomly rearranged, and then reassembled [810]. Figure 1 shows the original and rearranged images of two example scenes. The target was a small, shaded, neutral grey sphere located randomly within the scene and matched in mean luminance to its local surround. Target-detection performance was quantified by the discrimination index d′ from signaldetection theory [11], which is intended to reduce the effects of observer bias. The spatial variation of d′ over each scene was compared with the spatial variation of local colour properties quantified in the colour space CIECAM02 [12-14]. It was found that observers could detect the target in the rearranged images almost as well as in the original images. When combined, the local colour properties of lightness and the red-green and yellow-blue components of chroma accounted on average for 50% of the adjusted variation in observers’ detection performance when images were rearranged and for 55% when they were not. These levels are similar to levels reported elsewhere for the predictability of target detection and of eye fixations [7], with the remaining variation usually attributed to cognitive or neural factors.