Since all properties (anatomy, physiology and sensory processing mechanisms) of the human visual system are the product of evolution and postnatal development, it is reasonable to suppose that they have been configured to make the most effective use of the characteristics of the environment, according to the lifestyles of the individuals who exploit them. This optimisation has been shown most convincingly for fish and insects, but there are many information theory based studies, which point to a similar optimisation in humans and other primates. Here we aim to demonstrate this tenet experimentally.
In the first two studies, we conducted psychophysical experiments that required human observers to distinguish between achromatic pictures of slightly different faces or objects (i.e. shape discrimination). The stimuli were digitally modified to increase their deviation from the second-order statistics of natural images, making them increasingly “unnatural” in this respect. Performance was best when the stimuli had statistics similar to those found in the natural visual environment. We also explored performance when viewing stimuli monocularly, using foveal and peripheral vision. Our results show that performance in peripheral vision was best for slightly “blurred” morph sequences and that stimulus resizing (M-scaling) did not fully compensate for the deficiencies of peripheral vision. In a third study, a simple multi-resolution cortical model of the discrimination processes was shown to be capable of predicting the previous psychophysical results. A fourth study was carried out to explore the spatio-chromatic information content of natural scenes. We found that a particular subset of scenes (those containing red fruit on a background of green or yellow leaves) have properties that match those of the psychophysically measured achromatic and red-green chromatic contrast sensitivity funtions but not the blue-yellow chromatic contrast sensitivity functions, implying that the red-green opponent system is particularly well suited to the fruit/leaf discrimination task.