Scheuchenpflug R, 1999, "President Lincoln, I presume? How spatial quantisation influences face identification" Perception 28 ECVP Abstract Supplement
President Lincoln, I presume? How spatial quantisation influences face identification
Studies of face recognition with either spatially quantised (block averaged) or low-pass-filtered images reveal a sudden drop in identification rates at resolutions below 16 blocks per face. Performance above this limit is only slightly affected. This seems to indicate a low importance of local (small-scale) features for face identification. To test this hypothesis, I manipulated the position of the quantisation grid. Subjects were tested in a speeded recognition task with coarse quantised versions of 18 male faces. While there was the expected drop of identification performance below about 16 blocks per face, optimising grid position had an unexpectedly strong effect on identification rate. This could be tentatively explained by improvement of the localisation of relatively small features. Identification of pre-experimentally unknown faces may induce a reliance on global features. Therefore, in experiment 2 I used 52 pictures of celebrities which had to be identified at different levels of resolution. Identification performance did not reach an asymptote at 16 blocks per face. Possibly subjects rely on features represented only at finer scales of resolution when recognising familiar faces. Thus, previous experiments on face identification at different spatial resolutions seem to underestimate the influence of local (small-scale) features.
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 1999 Abstract Supplement (complete) size: 1478 Kb