Purchase this article with an account.
F Devinck, S Delord, K Knoblauch; Visual Cues Used by Human and Ideal Observers in the Perception of Luminance Patches . Invest. Ophthalmol. Vis. Sci. 2002;43(13):2914.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: We have previously shown (Devinck et al., ARVO, 2001) that observers detect a luminance patch based mainly on the summated luminance within the stimulus and not by its edges. Here, we ask if this strategy is also used in luminance discrimination. Methods: We used the classification image (CI) technique of Ahumada (1996) for a luminance discrimination task. The observer was continuously presented with a 3 deg square luminance increment on a 10 deg square background. A trial (400 msec) consisted of the random modulation about the mean of the luminance of each of the 64x64 pixels. In addition, on each trial, the 3 deg field was either incremented or decremented in luminance (d'=1). Observers classified trials as increments (I) or decrements(D) . Noise backgrounds were classified, according to the type of trial (I/D) and response (I/D), averaged and combined in a weighted sum (II+DI-ID-DD) to calculate the CI. CIs were also calculated based on the neural images of an ideal observer to investigate the influence of center/surround balance and receptive field size on the CI profile. Results: CIs for discrimination resembled those for detection, indicating summation of surface information and little contribution from edges. Over a wide range of receptive field sizes, an ideal observer, who based his decisions on the neural image from receptive fields with a balanced center/surround, yielded a CI based on edge information and required 6 times more contrast than a human for equivalent performance. A 20% imbalance in the center/surround interaction was sufficient to produce a CI based on the surface and to reduce the contrast requirements by a factor of four. Conclusion: Discrimination as with detection of a luminance patch is based on surface information because the pixels within the patch contributed to the observer's decisions. This information can be transmitted equally well through large or small receptive fields provided the center/surround interaction is unbalanced.
This PDF is available to Subscribers Only