Abstract
Figure-ground organization determines the shapes that we see at edges in images as is widely known by Rubin's faces-vase reversible image. Figure-ground assignment is influenced by a host of image-based cues as well as non-image factors such as attention. Recent work by Gillam and Grove (2011) has proposed a new cue of entropy which holds that regions containing high entropy (disorder) visual features are more likely to be perceived as background. This was tested in multi-partite images in which a rectangle was divided into columnar bars by straight edges and every other bar was filled with texture lines of several different random orientations (high entropy) or texture lines of all the same orientation (low entropy). The regions with high entropy lines were judged as in front less often than those with low entropy lines. We propose that these findings can be better explained as examples of the extant cue of edge-region grouping (Palmer & Brooks, 2008) rather than by a new entropy account. We conducted experiments to demonstrate this by varying the entropy of the texture lines within the regions but also independently varying the edge-region grouping relationship between the texture lines and the region borders. We found that edge-region grouping had a significant effect on perceived figure-ground organization whereas entropy did not. Our results suggest that edge-region grouping provides a better explanation of Gillam and Grove's findings than does the proposed entropy account.
Meeting abstract presented at VSS 2016