The imperfection of the Human Vision System

The Human eye-brain system is what we use for detecting, discriminating, naming, remembering and judging using color. It is not “perfect,” in the sense that it does not conform to a simple geometric model. This, to me, reflects the imperfection of the model, not the imperfection of the human system.

1 Like

I agree that it is the model’s job to reflect reality.
But apart from this: something matching “a simple geometric model” is a very unusual definition for the term “perfect”. Where does that come from?

But apart from this: something matching “a simple geometric model” is a very unusual definition for the term “perfect”. Where does that come from?

For context, this post was a response to a specific comment made by one of the speakers at the VisGuides 2018 workshop. I can’t recall exactly what the comment was.

There are some senses in which the human visual system appears to be optimal, or nearly so.
For example, the human eye is sensitive to single photons under appropriate conditions (there is an extended discussion of this in William Bialek’s textbook Biophysics: Searching for Principles). There is also an entire literature discussing the degree to which the receptive fields of visual neuroscience are determined by the statistics of ecologically relevant images. However, as I am neither a biophysicist nor a computational neuroscientist I cannot comment further.

In a trivial sense, color vision is clearly imperfect because we do not perceive the full spectrum of light: if we did, then the Wright and Guild color matching experiments wouldn’t be possible.
However, producing more cone types would have associated costs (for example, reduced visual acuity), and any serious discussion of optimality would have to consider this trade-off.

There is also an entire literature discussing the degree to which the receptive fields of visual neuroscience are determined by the statistics of ecologically relevant images

See, for example this article and the textbook Natural Image Statistics: A probabilistic approach to early computational vision.