Purchase this article with an account.
Wietske Zuiderbaan, Serge Dumoulin; The contrast response function is enhanced according to local subjective importance in natural images. Journal of Vision 2018;18(10):738. doi: 10.1167/18.10.738.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
INTRODUCTION Perception is based on the interaction between sensory information and our knowledge of the world, yet there is no consensus about how this interaction influences the neural signal in early visual areas. According to different inference theories, the knowledge-based perceptual hypothesis can either suppress or boost the sensory information represented in early visual areas. Suppressive or boosting effects are often referred to as predictive and effective coding respectively. Here we investigated the effect that the knowledge-based perceptual hypothesis has on the contrast response function in early visual cortex. METHODS We used 7T MRI to measure responses to viewing of natural images. For each image, we quantified the amount of RMS-contrast (sensory-driven) and subjective importance (knowledge-based). The subjective importance was based on the manual delineation of important regions of the image. We also measured the population receptive field (pRF) properties in early visual areas. We used the pRFs to quantify both the amount of RMS-contrast and the amount of subjective importance in the pRF. Combining this with the responses to natural images, we derived the contrast response function (CRF) for pRF locations with high and low subjective importance. RESULTS Based on the inherent variations of contrast in the natural images, we show that we can derive the CRF. This CRF is comparable to those reported in the literature. Furthermore, we show how the CRF is boosted in visual areas V1-V2-V3 for regions of high subjective importance as compared to low subjective importance. We found a similar boost of the CRF for figure vs ground in V1-V2 but not V3. DISCUSSION Subjective importance alters the CRF in early visual areas. We suggest that this alteration reflects the interaction between sensory information and our knowledge of the world. This interaction argues for effective coding and against most conventional implementations of predictive coding.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only