SNIP
When generating a rough profile, I often have to decide how
many samples to take before applying "auto complete".
Here are some scenarios:
1. After a sample, the indicators under the sliders have all
three rgb colors present. After a few such samples, I apply
"auto complete". But such cases are rare.
If you previously built a profile and saved it, you can use it on a
scan of the same type of film on the same scanner at the same ppi
setting. You can then use the actual image you want to treat,
optionally fine-tuning the profile, if it has featureless areas that
you could use for analysis. You can also use one (or more) other
image(s) from that roll to improve the profile you have so far. The
graininess is similar throughout the film. That is also useful if your
image has no featureless areas at all.
2. After a sample, some colors are missing in the indicators.
This is usually what happens. What should I do before
applying "auto complete"?
See above. You can use other images as well.
3. After a sample, one of the colors in the indicators have
broken lines. What does that mean and should I include this
sample before applying "auto complete"?
I'm not exactly sure what you mean with broken lines.
If you mean not all brightnesses are sampled, well that's what
autocomplete means; interpolate or even extrapolate to fill in the
missing parts. That would seldom be the case if you shot one target
image in the first place, and build a profile from that.
If on the other hand you mean that the line is irregular, you can
always change the measured values by hand with the sliders. It is
usually caused by some features (e.g. a smooth gradient) where you
should have measured a flat tone.
There are also warnings about color clipping. What does that
mean and how should I handle it?
It means that one or more channels will produce negative, or higher
than possible to encode values. It would mean that colors may be
changed when removing the noise, therefore NI warns that you better
not use that sample.
Above each slider there is a color coded %. % of what? How
should the color coding be interpreted?
The percentage is the positive/negative deviation from average noise
that was measured in the first "Rough noise" sample. It may well be
that dense film areas produce more than average noise levels, and
transparent film areas less. That will be reflected in the slider
settings. The colors of the percentages e.g. tell whether the sample
was aquired automatically, or changed by hand, or changed drastically
between sampling the same tone. The exact meanings are described in
the helpfile.
What is the difference between "auto complete" and "auto fine tune"?
Should both be applied, in what sequence?
Auto fine-tune automatically searches the entire image for featureless
areas, and automagically builds a profile without intervention.
However, as with anything automagic, the algorithm can fail to detect
featureless areas. Therefore you can additionally resample a few areas
to check if things change drastically. You can also use another image
to do that.
That may still leave a few brightnesses unsampled. Auto complete does
inter/extra-polation of those missing brightness values.
NI provides users with lots of features and information. It surely
would
help if we know what they mean and how to use them.
Well, they are described in the help file, but I understand it may be
a bit intimidation in the beginning.
They also have a support forum of their own, you can learn from
questions that others have asked in the past, or you can ask your own.
Of course some prior effort to find the answer in the help file will
speed up the resolution of issues, bacause you can tell what has been
tried and others can take it from there.
SNIP
NI recommends to sample only areas without real textures.
I try to follow that, often ending up with very few samples.
And so you should, but it can still lead to some misinterpretation of
what is noise and what is not.
SNIP
Knowing how to operate these previews is a great help. NI should put
this in their user's guide.
They are open for improvement suggestions.
You basically choose the channel you want to inspect, then
click/release/click release till you've got a feeling for what
changed. If it results in too much detail loss or not enough noise
reduction for a component, change the relevant amounts.
Once you gain some experience, you'll probably see that you will apply
more subtle amounts of reduction, rather than the default amounts.
That will result in an improved but not artifical looking image. Just
don't overdo it, and you'll be fine.
Bart
