Visual Perception and the Statistics of Natural Scenes
Neural systems adapt to the statistical structure in the inputs that they have to process. Well-known examples of such adaptation are the center-surround structure of the retinal ganglion receptive fields, the luminance and contrast adaptation in the retina, or the shape of the receptive fields in the visual cortex. In general, information theory teaches us that the design of optimal information processing systems depends on the input signals that they receive.
With regard to retinal information processing we have extensively studied the properties of natural visual scenes, such as the luminance and contrast spatial correlations, local luminance histograms, higher order and oriented edge statistics etc. We have assembled a large calibrated database of images, publically available here. ➤
We are currently using the higher order statistics of natural scenes to explore the perception of visual textures and shapes. (See publications.)