Purchase this article with an account.
Ilmari Kurki, Aapo Hyvärinen, Jussi Saarinen; Using classification images to reveal the critical features in global shape perception. Journal of Vision 2010;10(7):1174. doi: 10.1167/10.7.1174.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Radial frequency contours (RF; sinusoidally modulated circles) have been used to investigate how the visual information is integrated in global shape recognition. However, it is not clear, (A) if the integration is based on particular contour features (RF peaks or throughs; contour corners or sides) and (B) how similar processing of different RF shapes is. Here, classification images (CI), a psychophysical reverse-correlation technique was used to estimate the parts of the RF pattern that are critical for shape discrimination (RF pattern versus circle). Stimuli were composed of difference-of-Gaussian patches (center sd = 5.6 arc min, n=32) in an RF contour (r=1.5 deg). Position noise (jitter) along the radial axis was added. The standard RF contour had zero modulation amplitude (circle). The modulation amplitude of the test was adjusted to keep the proportion of correct detections at 75%. A one-interval shape discrimination task was used with 4-point rating scale. CIs were computed from the position noise. Both four-cycle RF4 and five-cycle RF5 patterns were tested. CIs show that both radial modulation peaks and throughs are used, suggesting that both contour sides and corners are about equally weighted in the shape recognition. The amplitude of the features across the contour length varies but is non-zero everywhere. This suggests that the detection is largely but not purely a global process. Especially, detection of the RF5 is based more on the features on top of the shape.
This PDF is available to Subscribers Only