The Wikipedia article on acutance claims that a slight amount of noise or grain can increase acutance. Cambridge in Color gives the following demonstration.
No noise:
With noise, with more acutance:
Why is this case? Is there a psychological or optical explanation for this phenomenon?
Unsharpness is the result of convolution with a kernel acting as a lowpass filter. The resulting image is characterized by a very low amount of high-frequency content (pixel colors changing at narrow distances). This is obviously desired for areas with smooth changes, like sky or flat surfaces. However, for items like bricks, we know that they contain a lot of pores and unevenness, and consequently should exhibit a high frequency of changes.
Adding noise (presumably mostly luminosity noise) brings the wall back to expectations since the kind of noise in images tends to be uncorrelated noise with the noise on each pixel independent from the noise on the next pixel. That gives noise an impact at all image frequencies, including the high image frequencies otherwise lacking in presence. That the additional "simulated" pores are at random locations not coinciding with where the "natural" pores blurred out would be is not apparent.
Of course, this trick does not work equally well for every kind of area.
It's an optical illusion. The brain is constantly seeking patterns and filling in detail where there is none.
The increase in acutance also occurs with patterns, such as canvas textures.
