Cite as:
Stollhoff R, 2010, "Representing facial information using Gabor wavelets" Perception 39 ECVP Abstract Supplement, page 127
Representing facial information using Gabor wavelets
R Stollhoff
In recognizing faces human observers extract facial information from different locations, spatial scales, and across different orientations. In experimental investigations these aspects have mostly been studied in isolation using techniques such as Bubbles (Gosselin and Schyns, 2001 Vision Research 41(17) 2261-2271), or bandpass-filtering of the Fourier spectrum. Here, a novel method to conduct an integrated investigation of these aspects is proposed and demonstrated in a simplified optimal observer model of facial information processing. Central to the method is a representation of face images by a family of Gabor wavelets modeled to mimic filter response profiles of simple cells in V1 [Lee, 1996 IEEE T Pattern Anal 18(10) 959-971]. In a simulation study using standardized face images, two properties of the wavelets were studied: Informativeness, calculated based on activation differences across faces, and robustness, calculated as the variability in activation across different levels of in-depth rotation. In general, wavelets displaying high informativeness and robustness are characterized by medium spatial frequencies, horizontal orientations - extracting vertical information, and clustering around diagnostic regions (eyes, mouth, ...). These results agree with similar findings in psychological experiments, and can provide a normative rationale for known specifics of human face processing.
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.]
