View Single Post
  #1  
Old September 12th 06, 10:36 PM posted to rec.photo.digital.zlr
Bill Again
external usenet poster
 
Posts: 27
Default Noise in digital picture

There has been some ongoing discussion about noise in digital pics and the
general assumption is that it is a "bad thing". Camera manufacturers work
hard to eliminate it, as discussed in the recent posts over the FZ30.
However, I came across a blog today which cites an academic paper suggesting
that noise can also be your friend. I have pasted the blog below, followed
by a link to the paper.

The comment about the possibility that Canon "build in about the right
amount of noise" is something that Panasonic might take on board.

Bill

PS The blogger is an architectural photographer based in the UK.
~~~~~~~~~~~~~~~
quote

This [see URL below] is an academic paper on the perception of sharpness in
digital photographs. I find this fascinating because it confirms a belief
that I have developed about why images scanned from film can still look
better than a digital image of apparently the same number of megapixels.

To my eye, digital images look unnaturally smooth as they lack the grain
which is exhibited by film. This paper suggests (among other things) that
adding a small amount of fake grain to a digital image can enhance its
perceived sharpness.

This is apparently quite an astounding discovery, because adding "noise"
using a tool like Photoshop, will in a strict sense reduce the amount of
correct information in the image. This is astounding to all but
psychologists of perception who know only too well that human perception is
a construct that overlays our preconceptions over the data we receive from
the outside world.

In this case, the human perceptual system appears to be fooled into thinking
that the extra random information which results from adding noise or
artificial grain to an image, is actually real information.

The amount of noise which needs to be added is actually rather subtle and
unfortunately this paper lacks enough detail about what kind or intensity of
noise was actually added. As a result, I can't see how it would be possible
to replicate their experiment accurately to confirm their results. I assume
that if this paper is eventually peer reviewed and published, that this
shortcoming will be addressed.

I've found that its possible to use Photoshop to add noise to achieve an
apparent improvement with a variety of digital file sizes from 6-39
Megapixels, and I have a theory that Canon engineers build in about the
right amount of noise (or rather let about the right amount through their
noise reduction algorithms) in the case of the 1DsII flagship camera.

Unfortunately the built in filters provided by Photoshop don't have the most
agreeable effect to my eye, I prefer proprietary third part software which
allows for more subtle and film grain like effects.

unquote

http://www.cis.rit.edu/people/facult...PDFs/PRO12.pdf