Fowlkes C C, Martin D R, Malik J, 2003, "On measuring the ecological validity of local figure - ground cues" Perception 32 ECVP Abstract Supplement
On measuring the ecological validity of local figure - ground cues
C C Fowlkes, D R Martin, J Malik
The significance of figure - ground organisation was first emphasised by Rubin [1921 Synsoplevede Figurer (Copenhagen: Gyldendalske)] who noted that size and surroundedness affect the likelihood that a surface is seen as figural. Stimuli designed by Kanizsa and Gerbino [1976, in Vision and Artifact Ed. M Henle (New York: Springer) pp 25 - 32] and Vecera et al (2002 Journal of Experimental Psychology: General 131 194 - 205) pointed respectively to the role of convexity and lower-region as cues. We expect that Gestalt laws make sense because they reflect the statistics of the natural world. If figural regions tend to be convex, the visual system can evolve to exploit that fact. Using annotated natural images, we are able to characterise ecological validity by directly measuring the frequency with which figural regions are smaller, below, and more convex than ground regions. We employed a set of 100 images that have each been hand-segmented into 5 to 30 disjoint regions corresponding to objects and surfaces. Each pair of abutting segments shared a contour. Subjects were shown an image and asked to identify which segment the contour 'belongs to'. We sampled 1200 boundary points from each image. The two segments associated with each boundary point were intersected with a local analysis window. Size was given by the number of pixels from each segment in the window. Lower-region was measured by the extent a vector normal to each segment boundary pointed upwards. Convexity was computed as the percentage of straight lines connecting point pairs in a segment that did not intersect the boundary. For each cue, we computed the logarithm of the ratio between the figure and ground measurements. We found that figural regions tended to be smaller, more convex, and occur below ground regions. The empirical distributions of feature ratios had a skewness of 0.52 for size, 0.29 for convexity, and 0.34 for lower region.
These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 2003 Abstract Supplement (complete) size: 1712 Kb