Humans can perceive the category of natural scenes accurately at a brief glance (Potter 1975), even when the scenes are presented as line drawings (Walther et al. 2011). What are the features underlying this ability? Biederman postulated that we use non-accidental properties, such as collinearity, curvature, or specific types of vertices, for the recognition of objects and their spatial relations (Biederman 1987). Practical tests of this model with real-world images have so far failed due to the challenge of extracting these non-accidental properties from photographic images. For our work we used line drawings that were generated by artists, who digitally traced the outlines in photographs of natural scenes. Having the exact coordinates of the artists’ pen strokes available allowed us to define non-accidental properties and other scene statistics using linear algebra. Specifically, we automatically extracted the distributions of contour length, curvature, orientation, angle between lines in intersections, as well as the counts of T, X, and Y junctions as defined by Biederman. We used these features to train a classifier to discriminate between six categories of natural scenes (beaches, city streets, forests, highways, mountains, and offices). The classifier could correctly identify the category for 86% of the line drawings in a left-out test set (chance: 16.7%). To assess the relevance of these features for human behavior, we compared the errors made by the classifier for the different types of features with the errors made by human participants in a six-alternative forced-choice categorization task of briefly presented and masked line drawings. Correlations of the off-diagonal elements of the confusion matrices were significant at p<0.05 for intersection angles (r=0.44), junction type (r=0.41) and contour curvature (r=0.37). This match between non-accidental properties and human behavior serves as experimental confirmation of the importance of these features for the perception of natural scenes.
Meeting abstract presented at VSS 2012