Purchase this article with an account.
Felix Wichmann, Nicole Eichert, Heiko Schott; An Image-Based Multi-Channel Model for Light Adaptation. Journal of Vision 2016;16(12):566. doi: 10.1167/16.12.566.
Download citation file:
© 2017 Association for Research in Vision and Ophthalmology.
The human visual system is sensitive to luminances ranging from 10-6 to 108 cd/sqm (Hood & Finkelstein, 1986, Handbook of Perception and Human Performance, Vol 1). Given the more limited dynamic range of the photoreceptors and subsequent neurons, effective light adaptation must thus be an essential property of the visual system. In an important study Kortum & Geisler (1995, Vision Research) measured contrast increment thresholds for increment-Gabor probes on flashed backgrounds in the presence of steady-state backgrounds, exploring how the spatial frequency of the increment-Gabor affects thresholds, i.e. how light adaptation and spatial vision interact. In addition to their experiments, Kortum & Geisler presented a successful model in which the incoming signal undergoes multiplicative and subtractive adapting stages followed by non-linear transduction and late noise. Here we significantly expand and modify their model: First, our model is image-based and thus accepts any image as input, whereas the original model is only applicable to the putative scalar activations of sine wave stimuli. Second, our model has a single set of parameters for the multiplicative and subtractive adaptation stages, followed by a multi-scale pyramid decomposition (Simoncelli & Freeman, 1995, IEEE International Conference on Image Processing, Vol. III). The spatial frequency dependence of the thresholds is modelled via the DC components' selective influence on the variance of the late noise in the decision stage of the model. In the original model by Kortum & Geisler the parameters of the multiplicative and subtractive adaptation stages are all spatial frequency dependent, which is problematic if one believes adaptation to happen very early in the visual system, before the signal is split into separate spatial frequency channels. Our image-based multi-channel light adaptation model not only accounts well for the data of Kortum & Geisler (1995), but in addition captures, for example, the effects of test patches of different size (Geisler, 1979, Vision Research).
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only