Image Processing Reference
Cutting Out the Noise
Can't See the Forest for the Trees?
Removing noise from your content is akin to film restoration; whether it is worthwhile
depends on the value of the footage. Removing the noise will help reduce the work the
compressor has to do to attain a low bit rate.
Weird Artifacts and Problems
There are a few nasty “gotchas” that may present themselves when you are processing
source footage ready for compression. They have been known and identified for years.
Most of them are inherited from the analog domain that older footage is derived from.
I will enumerate them here and describe what causes them and whether anything can be
done to alleviate them.
Vision Problems and Imaginary Artifacts
Some visual artifacts are imaginary, and others are due to defects in the viewer's vision
such as color blindness or astigmatism. A few are consequences of the way that the human
eye works and therefore cannot be compensated for or corrected.
Noise in the Source Image
The noise in any source material is going to be seen by the compression engine as differ-
ence information. There is a certain level of random white noise that creates a twinkling
effect in the intensity of an analog video signal. This is apparent when you look closely at
a locked-down shot of a studio. The background is supposed to be unchanging, but on a
per-pixel basis, it twinkles all the time. Removing this noise before applying the compres-
sion is well worth the effort, as it reduces the amount of work that the compressor will
have to do.
Preserving hard edges but softening a large area of continuous tone in the luma
channel should yield a useful improvement. Ideally, the smoothing should take place in