[Tig] DVNR settings theory and practice -- advice needed
Tue Oct 9 20:33:02 BST 2001
First, we must note that there are two types of DVNR Noise Reducers for both
High Definition and Standard Definition processing. All of Digital Vision's
post-production products are modular and start with a Base Frame.
Processing option cards are then added to the frame. In SDTV Digital Vision
manufactured an NR (older Noise Reducer) and ANR (Adaptive Noise Reduction).
The ANR incorporates the PHAME? motion algorithm and is far superior to our
older NR product.
"... tape-to-tape color correction and noise reduction were done (daVinci
2K, DVNR 1000....DVNR settings (14 Y, 14 C) were used. (These were used
often. No Y or C prefiltering was used, as the colorist felt that they are
to be avoided.)"
The key to quality noise reduction is the systems ability to detect motion
and compensate for this motion by limiting the amount of artifacting caused
by temporal processing. To augment the systems power, the ANR permits the
operator to control the Luminance (Y) and Chroma (C) channels for both the
recursive and spatial filters independently. The Y & C strength settings
you noted "(14 Y, 14 C)" are very strong settings for the ANR. The
Y-strength & C-strength settings are specific for the recursive or temporal
filters. Because we predominately see the luminance signal and chroma
information "sits behind" the luminance channel, we must be careful as to
the level of our Y channel. Due to the nature of the video signal, we can
"get away with" more processing in chroma channel but please be aware that
we are certainly "getting away with" this increase in processing.
Important Note - The "Adaptive" process of the ANR uses PHAME? to adapt to
an image's motion and shift processing of moving images from recursive
filters to the spatial or Pre-Filters. In short - the Y-STR and C-STR will
effect only parts of the image that remain static while the Y-Pre and C-Pre
effect portions of the image that move.
Therefore, if the purpose of Adaptive Grain Reduction is to eliminate or
reduce recursive motion artifacts by "turning off" the recursive filters
based on information from the PHAME? process, one may ask, shouldn't we be
concerned about the Luminance settings in our Y-PreFilter. (The Pre-Filter
processes movement only) The answer is, yes. Note, a spatial filter does
not create temporal lag. Only a temporal filter can create temporal lag.
The spatial filter does, however, have specific artifacts that an operator
The key to quality ANR is to control the level of Luminance processing in
both the Recursive and Spatial filters. A "Normal" setting would be
Y-Strength = 2, C-Strength = 4, Y-Pre = 0, C-Pre = 3. Note, that the ANR
strengths in SDTV only is Linear. The ANR settings in HD are Logarithmic
(Subtle settings on the low end and increasingly stronger settings above 5).
They key to the above setting is the Low Y-Strength setting and the Zero for
Y-Pre-filtering. This will help avoid the "...a slowly-moving film of grain
superimposed over the image." This is indicative of a strong recursive
setting of the Y & C Strength filters - temporal lag. It is especially
important to avoid temporal lag when pre-processing for compression. Over
processed images (temporal lag) can be equally difficult for encoders as the
original grain one is trying to reduce. HOWEVER, this is why I stress the
importance of the Digital Vision Brickwall Filter for all image restoration
and pre-processing. I could spend another hour explaining the importance of
the BWF, however, I strongly suggest you investigate the BWF and ANR (or
AGR-IV) combo for pre-processing prior to compression. (Note - this is
commonly found in our BitSaver501 or DVNR500-PP package)
With regards to the new AGR-IV. For the first time in seven years Digital
Vision has redesign its marquee product and is currently BETA testing the
AGR-IV in the United States. The HD AGR-IV is a GPB based processing
option, resolution independent up to 4K and processes both RGB and YUV -
(Dependant on the Base Frame I/O). Digital Vision has resolved several of
the older ANR's limitations and augmented and redesigned the entire
processing engine and PHAME? algorithm. So far, the results of our efforts
have been widely praised. We will be demoing the new AGR-IV this Wednesday
and Thursday October 10 & 11th at our Open House in North Hollywood. Feel
free to contact me directly for a personal demo of the system or feel free
to stop by between 12:00PM - 8:00PM Wednesday or 10:00AM - 6:00PM Thursday.
I hope this response empowers everyone to use the system more wisely and
avoid any costly mistakes. As stated, "It's a great box, but can be
dangerous if overused."
US & Mexico Sales Manager
Digital Vision (US) Inc.
4605 Lankershim Blvd. Suite 700
North Hollywood, CA 91602
Please note that the Luminance and Chroma Adaptation channels will also
assist in optimizing the processing for any image source.
From: JSnopes at aol.com [mailto:JSnopes at aol.com]
Sent: Saturday, October 06, 2001 1:33 PM
To: tig at emsh.calarts.edu; tig at alegria.com
Subject: [Tig] DVNR settings theory and practice -- advice needed
I need a bit of advice re DVNR settings.
I'm working on remastering a documentary from the Seventies, with very
pretty 7242 color, and quite a bit of nice grain (much of it was pushed
The original transfer was a flat transfer on a Spirit from the 16mm
interneg (A&Bs weren't practical to work from) to D5HD24P. This was
downconverted (via Teranex) to Digibeta. (The D5HD master is being
stored, not needed now.)
Thousands of dust fixes were done in an Avid (uncompressed).
Then tape-to-tape color correction and noise reduction were done (daVinci
2K, DVNR 1000).
Color correction is beautiful. And the noise reduction looks (mostly) ok
when viewed on a 19" Sony BVM20F1, as it was in the telecine suite.
However, when the same tape is viewed on a calibrated 29" Sony monitor
(component in, from SDI via AJA D to A) three things are apparent.
First, and most noticable: there's a sort of grainy veiling over the
image, mostly on scenes where higher DVNR settings (14 Y, 14 C) were
used. (These were used often. No Y or C prefiltering was used, as the
colorist felt that they are to be avoided.) It looks like a
slowly-moving film of grain superimposed over the image.
Second, (mostly) on scenes with extreme contrast ratios, with, say, a
side-lit face, when the face moves, the bright side of the face and the
dark side of the face will move separately, as if they are shifting
tectonic plates -- the dark side's movement lagging a couple frames
behind the bright side. This can be quite annoying. Again, less visible
on the 19" monitor, but there if you are attuned to it.
Third, no aperture correction was used. This may have been a
misunderstanding, as this colorist rarely does tape-to-tape, but
regularly uses aperture correction on their Spirit-to-tape work. This
results in a slightly mushy looking image -- when I run it through our
DVNR for aperture correction only (a 2 or 3 on H and V settings) it looks
(to me) much better.
So, what I'm really wondering is:
What are the usual limits (i.e. highest settings you'd use) on a DVNR.
To me, from some rough tests, it seemed like a 9 Y and a 9 C was about
the limit (on this grainy material). Is prefiltering something to be
used or avoided?
What sort of aperture correction (standard, not advanced) do most of you
like (both H and V) for material that you've noise-reduced?
Any other advice as to the use and overuse of DVNR? It's a great box,
but can be dangerous if overused. (BTW, purpose of this master is DVD
Anyone used the new DVNR card yet (it's being beta tested now, I hear)?
All advice greatly appreciated.
Jeff "noisy, and grainy too" Kreines
Tig mailing list
Tig at tig.alegria.com
More information about the Tig