Cite as:
Galperin H, Lisitsyn D, Bex P J, Fiser J, 2010, "Feature classification in natural scenes" Perception 39 ECVP Abstract Supplement, page 62

Feature classification in natural scenes

H Galperin, D Lisitsyn, P J Bex, J Fiser

In a new, generalized classification image framework, we performed a post hoc metaanalysis across seven psychophysical experiments to determine the underlying features utilized during object identification. Stimuli consisted of images of everyday objects constructed from 700 Gabor patches across three spatial frequencies. Signal Gabor elements were assigned positions and orientations defined by the object image, while noise Gabor elements were assigned random positions and orientations. Signal Gabors could be selected independently from image areas composed of smooth edges, corners or randomly. Gabor interelement pairings for all images from 21 000 trials were analyzed. We obtained the distance, element orientation, and contour curvature difference between all possible element pairings, yielding element pair distributions for signal and noise elements present on correct or incorrect trials. These distributions were analyzed to determine which element pairings significantly modulated performance. Classification images showed peaks for nearby Gabor pairings whose joint orientations were either similar (smooth contours) or orthogonal (corners and end stopped). These results challenge previous claims that particular orientation structures, such as edges or junctions, are principal features for object recognition, but instead show that the visual system employs a generalized weighted combination of features in object recognition tasks.

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 2010 Abstract Supplement (complete) size: 3078 Kb

[Publisher's note: The abstracts in this year's ECVP supplement have been published with virtually no copy editing by Pion, thus the standards of grammar and style may not match those of regular Perception articles.]