The human visual system is optimised for processing the spatial information in natural visual images.

Area of Science:

• Visual Science
• Neuroscience
• Evolutionary Biology

Background:

• Visual systems are believed to adapt to their environment through evolution and experience.
• This principle is well-established in fish and insects, but less so for mammalian vision.
• Previous evidence for mammalian vision has been largely theoretical.

Purpose of the Study:

• To empirically test if human spatial vision is optimized for natural visual environments.
• To investigate the role of image statistics in human visual perception.
• To provide direct evidence supporting the adaptation of mammalian visual systems.

Main Methods:

• Human observers performed discrimination tasks using images of faces and objects.
• Images were generated using a morphing technique to control for quantifiable changes.
• The independent variable measured deviation from natural scenes using Fourier composition (second-order statistics).

Main Results:

• Observer performance was highest for images with natural second-order spatial statistics.
• Performance degraded significantly as images became less natural.
• A simple contrast coding model, based on mammalian visual cortex simple cells, explained the results.

Conclusions:

• Human spatial vision is empirically demonstrated to be optimized for the second-order statistics of the natural optical environment.
• These findings provide direct support for the evolutionary adaptation of mammalian visual systems.
• The study links visual perception performance to the statistical properties of natural scenes.