View Full Version : What is the deal of grain retention


Chengbin
29th May 2009, 04:01
I should have asked this question a long time ago. Lurking around the forum, and I often hear goals of retaining grain.

Why in the world would you want to retain grain?!?

I thought grain was an unwanted artifact from video cameras. Why would you want that?

In real life the world isn't filled with grain, so it is not for realism.

For me MVDegrain 3 is one of the best feature of encoding.

If it is an artifact, and it is not for realism, why would you want to waste huge amount bits on grain??

Adub
29th May 2009, 05:06
Grain is also an "Ambiance" factor. A mark of "classic" film is that of natural film grain. Film grain has been in movies for ever, and some (lots actually) like the affect that it provides. Reminds them of old times or something. In addition it prevents banding by providing dither, and gives the illusion of detail as well.

I filter my film grain heavily, but just so that there is only a sprinkling left. I don't eliminate it entirely. But that's me.

*.mp4 guy
29th May 2009, 05:22
... there is also the small problem that separating grain from detail perfectly is impossible.

Blue_MiSfit
29th May 2009, 06:06
... there is also the small problem that separating grain from detail perfectly is impossible.


Exactly. Removing "only" the film grain is almost impossible in all cases.

If we're talking about CCD / CMOS sensor noise (aka "grain") then yes, this is highly objectionable and can be removed in many cases - particularly chroma noise. That ugly blue and red noise you see on high ISO digital camera images is what I'm referencing here :)

Grain is a part of all true film. Sometimes I find it best to just keep as-is, and sometimes I filter it. Very rarely do I try to remove "all" grain, because this usually results in an ugly, over-filtered image. It also often creates banding in the unfortunate 8 bit YV12 color-space we get to work with :( In fact, in most cases I add artificial grain after processing images for this very reason.

Still, to each his own. MDegrain/2/3 are all amazing filters, and there's some seriously good hardware out there in the pro sector as well.

~MiSfit

Daiz
29th May 2009, 17:02
I personally like the look of grain myself, but too much is still too much, though. And even if I didn't like it, I'd still prefer it over (over)smooth video with ugly banding.

zn
29th May 2009, 18:40
Digital cameras like RED (http://www.red.com/) is way to go (http://www.reduser.net/), hope they will trash their old cassete cameras soon.

And I hate when they digitally add grain, unless it is needed for effects like in "300" movie.

They can add "grain" or "black & white" features to disc players, for customers who need it.

Tagert
29th May 2009, 22:30
I love grain :)

Mtz
29th May 2009, 22:35
Grain was invented for HD to hide the problems of imperfect/immature Plasma/LCD and HD encodings.
I hate too much grain. I hate grain.

enjoy,
Mtz

*.mp4 guy
30th May 2009, 07:56
Exactly. Removing "only" the film grain is almost impossible in all cases.

If we're talking about CCD / CMOS sensor noise (aka "grain") then yes, this is highly objectionable and can be removed in many cases - particularly chroma noise. That ugly blue and red noise you see on high ISO digital camera images is what I'm referencing here :)

High ISO digital often looks horrid, but you still have to be careful about how you handle it, or you will get a really washed out look. Also, even at low 100-400 ISO's, you will have noise, and removing it would be just as destructive as removing grain from film.

Grain was invented for HD to hide the problems of imperfect/immature Plasma/LCD and HD encodings.

no

Digital cameras like RED (http://www.red.com/) is way to go (http://www.reduser.net/), hope they will trash their old cassete cameras soon.

Digital cameras have "grain" too, often to a far larger extent then large format low grain film.

MfA
30th May 2009, 10:22
Digital cameras have "grain" too, often to a far larger extent then large format low grain film.
Non spatially correlated noise is easier to remove though.

*.mp4 guy
30th May 2009, 20:12
In practice, it is spatially correlated by the complex debayer filters (all high resolution digital cameras use bayer AFAIK). Also, theoretically, the fact that the ccd is partitioned into squares and not points will introduce frequency correlation aswell. Also, film grain should have a simple gaussian correlation, which is not exactly difficult to deal with. Furthermore, none of the denoising methods I would consider actually using would be significantly negatively affected by a simple positive spatial correlation.

MfA
30th May 2009, 22:52
The sample footprint is determined more by the anti-aliasing filter than the sensor shape.

There is nothing simple about spatial correlation, it's the only real distuingishing feature of signal compared to noise you have to work with in spatial denoising ... the only methods which wouldn't be negatively affected are purely temporal.

LoRd_MuldeR
30th May 2009, 23:31
I think the primary reason why we need "grain retention" is because of this: If a decoder cannot retain the grain in a movie, then the result will not be a perfectly degrained video, but an ugly blurred mess! So if you want a good quality encode, you must either use high quality pre-processing that removes most of the grain without generating additional artifacts (e.g. MVDegrain) before encoding -or- you must use an encoder that retains the grain in a decent way. But high quality pre-processing is slow, really slow. Plus there are far too many thing that can go wrong with bad noise reduction, as seen in many releases! And as mentioned before, not everybody likes degrained video. The grain gives the film a "natural" and more detailed look. Strong denoising creates an "artificial" look. That's why I'm usually not a big fan of denoising...

Chengbin
31st May 2009, 02:02
The grain gives the film a "natural" and more detailed look. Strong denoising creates an "artificial" look. That's why I'm usually not a big fan of denoising...

I'm not sure if I agree with grain give a video a more natural look. People don't have grain on their faces. That's why I love MVDegrain 3. It restores the natural look of the video.

LoRd_MuldeR
31st May 2009, 02:11
I'm not sure if I agree with grain give a video a more natural look. People don't have grain on their faces. That's why I love MVDegrain 3. It restores the natural look of the video.

Different people, different preferences :D

IMO quite often you see faces that have been degrained much too strong and thus look unnatural. More like dolls than humans ;)

(That maybe doesn't happen with a good degrain tool, such as the one from MVTools)

Sharktooth
31st May 2009, 02:17
Grain was invented for HD to hide the problems of imperfect/immature Plasma/LCD and HD encodings.
I hate too much grain. I hate grain.

enjoy,
Mtz
complete nonsense. grain exists since the first life form with a visual system called "eye" appeared on earth.

Soulhunter
31st May 2009, 11:46
complete nonsense. grain exists since the first life form with a visual system called "eye" appeared on earth.
True, I can clearly see noise in front of my eyes, especially when its dark! ^^;

scharfis_brain
31st May 2009, 11:59
And since there is grain being produced by the eye our neural network called 'brain' is able to filter it out temporally if you stare for long enough on a object ;)

Sharktooth
31st May 2009, 13:58
*** BUT ***
guess what? videos are not a stable objects (images)...

scharfis_brain
31st May 2009, 14:04
That's where your brain's object tracking starts to work ;)

Chengbin
31st May 2009, 14:33
True, I can clearly see noise in front of my eyes, especially when its dark! ^^;

That's right! I can see grain/noise when it is dark. What's up with that?

Sharktooth
31st May 2009, 14:43
that's due to "rods and cones" aka the photoreceptors cells... we see "pixels" too and we have different receptors for luma and chroma...
human eyes are not much different from a high quality digital camera... so they also "see" grain...
post processing is done by our brain though, but still we see grain...

Soulhunter
31st May 2009, 20:55
And since there is grain being produced by the eye our neural network called 'brain' is able to filter it out temporally if you stare for long enough on a object ;)
Btw, funny random fact: It seems I perceive blue color with a delay when its dark. For example, when I slowly move a cigarette in front of my eyes in complete darknes, I clearly see a blue "ghost (http://img5.imageshack.us/img5/6687/212a.png)" of the fervor following (1-3 cm distance) the actual one! Does the eye also change its exposure time for blue light in darkness or so? O.o

scharfis_brain
31st May 2009, 23:03
If I look at the green lit number of my radio clock in the darkness, I see the bright number w/o delay and the rest of the room with delay.

for sure the eye or brain integrates the low light areas slower than bright areas that's where the delay comes from.

Midzuki
1st June 2009, 00:37
human eyes are not much different from a high quality digital camera...

Which one was invented first? :)

Lyris
1st June 2009, 00:40
I should have asked this question a long time ago. Lurking around the forum, and I often hear goals of retaining grain.

Why in the world would you want to retain grain?!?

I thought grain was an unwanted artifact from video cameras. Why would you want that?
No: Grain is part of the FILM process. NOISE is a usually unwanted artefact from video cameras.

Grain also increases the perceived detail.

Whoever said realism was the intention anyway?

*.mp4 guy
1st June 2009, 05:32
If you pay attention to your visions characteristics, it becomes clear that the brain does a lot of processing before we actually "see" anything.

Extrapolating from rod/cone density, inherent noise constraints, and the dynamic range and resolution of the HVS, its clear that some fairly impressive voodoo happens at some point.

Gokumon
3rd June 2009, 03:00
I should have asked this question a long time ago. Lurking around the forum, and I often hear goals of retaining grain.

Why in the world would you want to retain grain?!?

Because in most cases when people try to degrain movies you end up with people with waxy looking faces and smearing in motion. Just look at the degrained Blu-Ray movies that have come out to see this.

Mug Funky
3rd June 2009, 09:26
OMG, i laughed several times in this thread.

What is the deal of grain filtering?

if a codec can preserve grain, imagine what details it can keep.

when the grain's gone, the encoding/prefiltering process is no longer transparent (which it should always strive to be, assuming you're working with a source that has had some form of direction, where we can assume it looks the way it's supposed to).

digital "grain" is dark current from CCD or CMOS sensors this comes in the dark areas obviously because it's an artefact of the sensor's natural noise floor. heat increases it, as well as interference from any number of things.

film grain is usually most apparent in the dark areas too, because when the negative hasn't got sufficient exposure, only a few silver grains will be chemically processed, and they'll show up as big grains (the actual silver is bleached out, the "grain" left behind is a spot of dye attached to where the silver was). you'll find the current state of the art in film is almost completely grainless even at high sensitivities, and if you're looking at a standard def transfer of a 35mm film, there'll be almost no grain at all.

"300" didn't have grain added to it - it was shot with very flat light (so hardly any contrast), and had the contrast cranked up to 11 in the colour grade. this pushed out the grain that was inherent in the film.

i'll only filter grain at all if (a) a client requests it because they have a few shots that are sticking out among the rest, or (b) my target bandwidth is so low that filtering noise is the lesser of two evils (the other evil is blocks/smears/bands/compression artefacts). with that said, i tend to work with pretty good sources - there'll be no added noise from compression or transmission, as it's all off a camera tape, master, or roll of neg.

onesloth
3rd June 2009, 18:38
What is the deal of grain filtering?
...
i'll only filter grain at all if ... my target bandwidth is so low that filtering noise is the lesser of two evils
^the deal. (for personal backups)

Mtz
3rd June 2009, 21:00
grain exists since the first life form with a visual system called "eye" appeared on earth.
When my eye is looking to the sky, I see "pure sky", no grain. Also when I'm looking to my car or at the faces of the people. No grain, no red faces, just eveything natural.
This is what I hope from a movie.
When a director wants some thing in his movie, OK, is his choice, but I don't want to see grain in ALL movies. Also I see no grain in TV broadcasting. But as I wrote in other post, they neeed to hide some problems of HD.
Now I'm waiting Pixar movie with grain.

enjoy,
Mtz

Sharktooth
5th June 2009, 02:57
no one said that. but an encoder should be able to "keep" the grain if it's in the source.

R3Z
20th July 2009, 09:37
Grain is beautiful, i like to think of it as the perfection of imperfection.

Heavilly filtered moving pictures just look devoid of all life and character. Cant imagine movies without it to be honest.

Ghitulescu
20th July 2009, 09:54
Grain is beautiful, i like to think of it as the perfection of imperfection.

Heavilly filtered moving pictures just look devoid of all life and character. Cant imagine movies without it to be honest.

I've seen yesteday "Funny Face" in 720p on ARTE HD. No grain at all, perfect image (great restoration). I own some DVDs with film grain, I don't like them from the image point of view. The grain issue it's a matter of taste. Just because one is used to something, it doesn't mean it should be like this.

On the other hand, trying to remove the noise grain that yield detail loss, that's a completely different story. It's bad!

Lyris
20th July 2009, 16:45
Even if the grain reduction doesn't destroy actual detail, it will lower the perceived detail, which is nearly as bad.

kolak
20th July 2009, 20:11
Film grain is nothing more than inperfection in recording process. Film grain became a digital nosie now and both are unwanted side effects of recording process. People get use to grain so much, that they think it's smething wanted. It's not! There are new digital cameras coming (like RED) and soon picture will be perfectly clean as it's in the reality seen by eye. Is anyone seeing any grain in the real life?
Because it was immposible to avoid it directors started using it as a effect and that's why people get use to it so much and think that it's something "real", which can't be avoided. It can- all we need is much bigger sensor and better technology, which is coming.
Film grain is the same side effect (inperfection in recording process/device) as digital noise with modern cameras or even home webcams- it doesn't exists in real life :) (at least is no seen by eye, so we don't need to "create" it during recording- or we can, but only to add artistic value).

Andrew

Ghitulescu
20th July 2009, 21:27
Film grain is nothing more than inperfection in recording process. Film grain became a digital nosie now and both are unwanted side effects of recording process. People get use to grain so much, that they think it's smething wanted. It's not! There are new digital cameras coming (like RED) and soon picture will be perfectly clean as it's in the reality seen by eye. Is anyone seeing any grain in the real life?
Because it was immposible to avoid it directors started using it as a effect and that's why people get use to it so much and think that it's something "real", which can't be avoided. It can- all we need is much bigger sensor and better technology, which is coming.
Film grain is the same side effect (inperfection in recording process/device) as digital noise with modern cameras or even home webcams- it doesn't exists in real life :) (at least is no seen by eye, so we don;t need to "create" it during recording- or we can, but only to add artistic value).

Andrew

My words :)

Soulhunter
20th July 2009, 22:03
Is anyone seeing any grain in the real life?
Yes, we do... Noise...

See some posts above!

Ghitulescu
20th July 2009, 22:33
Yes, we do... Noise...

See some posts above!

I don't see noise in real life. I see noise on TV and DVDs, but again not in real life. If one see grains while looking at a tree, a lake, a meadow, it should go to a doctor - has a visual problem.

*.mp4 guy
21st July 2009, 02:12
There is noise in human vision, it doesn't look much like film grain, but if you don't see it you aren't paying attention.

Anyway, most people here aren't arguing for grain being positive in general, its just that keeping it is the lesser of two evils. On the other hand, arguing for DNR is like arguing that we should remove the highly visible brush strokes from Van Gogh's paintings, it completely misses the point.

Ghitulescu
21st July 2009, 07:49
There is noise in human vision, it doesn't look much like film grain, but if you don't see it you aren't paying attention.

Anyway, most people here aren't arguing for grain being positive in general, its just that keeping it is the lesser of two evils. On the other hand, arguing for DNR is like arguing that we should remove the highly visible brush strokes from Van Gogh's paintings, it completely misses the point.

Then you should know that there are movies wherein the grains are added, just to keep the kino feeling. BluRay specifications allows the addition of artificial noise.

About the removal of noise that also kills the details (I've seen a Bollywood movie, the action was in a desert, the tiny sand waves were completely removed in the noise removal process, so it looked like a 3D anime). I fully agree with all that say that grain should be kept if its removal adversely affects the image quality. But it is possible (just expensive) to remove it without too much impact. What I do not like is that grain is artificially added, or that the studios (that charge a lot for their movies and managed to get yet again a prolongation of their rights) fool the people that grain is good. More money and rights for less work.

Mug Funky
21st July 2009, 08:15
i've never heard a director, while on a press junket promoting his film, say "yeah, we think you'll all really like this because we added a bunch of grain". studios don't care what the director requests or doesn't request in post, and there's certainly no International Grain Conspiracy by the big 3.

name 1 movie where you know for a fact grain has been deliberately added. i dare you. i know of 1 short film, and that was only to hide the fact they were cutting from 500T to 50D (the higher the number, the grainier. 50D is almost perfect, and therefore the cut between the film stocks became quite jarring).

now, that said, i don't agree with some clients adding grain to RED footage. it already has enough (that's right, i said it. it's grainy. deal with it).

human vision most certainly has grain. your brain's probably filtering it out... a clear blue sky will be made up of infinitesimal noisy bits when you look up at it. what's more, human vision has a significant "afterglow" effect that makes things even more weird looking. photodiodes have been able to outperform human vision for years now. the human eye can detect a flash of about 6 individual photons IIRC, which is insanely sensitive. but a PEC in the right mode can detect single photons quite reliably. there's the occasional misfire from shot noise, but it wont miss a photon. CCDs are able to do this as well in some cases, but with dark-current and shot noise problems.

one thing the human eye does NOT do is remove details willy-nilly in order to present the brain with a "cleaner" image. so philosophically i don't think our encodes should do this either... unless bandwidth or artistic intent say otherwise (obviously you don't want noisy footage in a cosmetics commercial...)

*.mp4 guy
21st July 2009, 08:16
Its only possible to remove noise without significant detail loss when the detail level is fairly low. Also, there is nothing wrong with having noise present purposefully for artistic reasons. The blu ray specifications are intended to allow adding noise that was originally in the source back after it has been lost due to compression, this is useful for obvious reasons; but I've never actually seen it used myself. There are no noiseless cameras, and thats unlikely to change, I seriously doubt that any studios are adding grain, many of them are doing exactly the opposite, and badly.

Mug Funky
21st July 2009, 08:40
There are no noiseless cameras, and thats unlikely to change

http://www.panavisionimaging.com/imagers_DMAX.htm

i'd say it'll still have some noise... but it'll be pushing pretty close to the 10-bit floor i think.

can't wait to see this bugger unleashed.

Ghitulescu
21st July 2009, 08:56
There is noise in human vision, it doesn't look much like film grain, but if you don't see it you aren't paying attention.

Of course there is noise in human vision, like there is noise in human hear. But the grain in movies is much much much greater. It's a defect, and people seems to get used to it and don't ask questions. Its removal would take away the "personality" (someone said before) only if it's incorrectly done.
It's like getting used to the saccadated noise of the shutter (of the projector). Would you like to hear it? Maybe you do, since snapshot cameras do this noise, to suggest you the old ambiance of an old SLR.
Remember what Chaplin said? The advent of sound would kill the movie?

smok3
21st July 2009, 09:23
examples;
a. if you shoot on 35 film, and then do some sort of compositing, you will have to find a way to add matching grain to the composited elements
b. if film is master, then you want the video to look similar (keep some grain)
c. if workflow is 100% digital, then grain/noise adds some personality (to certain scenes), but it is not needed
d, it should be easy to test the same scene with different amount of grain and see what it means to a person (how much can we abuse the imagination that will fill the grainy gaps)
e. if you check autodesk lustre manual (available online), you will see it mostly talks about grain removal.
http://usa.autodesk.com/adsk/servlet/index?siteID=123112&id=11992855&linkID=10809889

Astrophizz
21st July 2009, 10:44
If human vision already has noise there's no need to emulate it in a video as just watching the video with your eyes will add that noise.

smok3
21st July 2009, 10:48
yes, but this is a practical world (not theoretical) one valid example for video would be:
1. denoise video (all kind of camera noises types)
2. add some grain (which is sexier that noise and it will mask the remaining noise and denoising artifacts)

Ghitulescu
21st July 2009, 10:53
Film grain is a defect that was magically transformed into a feature. This feature even has supporters. Maybe the software developers know this already: it's not a bug, it's a feature!
Why the professional photographs used glass matrixes instead of regular 35 films? Because at their size (10") the emulsion grain was less evident when enlarging the photo for a magazine or a poster. Why using less ISO (under 100, like 60 or even less)? Because it was easier to manufacture fine grain emulsions at lower ISO (the higher the ISO the higher the grain).
Does not sense a professional photograph the artistic value of the grain? I very doubt it...

Ghitulescu
21st July 2009, 10:56
yes, but this is a practical world (not theoretical) one valid example for video would be:
1. denoise video (all kind of camera noises types)
2. add some grain (which is sexier that noise and it will mask the remaining noise and denoising artifacts)

The denoising tools a private may have do not match the tools the producers [should] have. If I do a denoising using my knowledge, yes, it would be better to leave the grain where it is. But I'm talking about the movie companies. They should be able to obtain fine granulation movies...

smok3
21st July 2009, 10:58
with photography is about making the film size as small as possible (costs) with as little grain as possible, but that wont stop me using bw 400ASA films of course, so to each its own... (of course i don't give a flying f*** about wedding photography...), that said i wouldn't mind having one of those hasselblads that were left on moon.

They should be able to obtain fine granulation movies...
things will change drastically when there will be no more film of course.

The denoising tools a private may have do not match the tools the producers [should] have.
mistifing professions is probably the reason for this thread (don't do it, Lucas is not a god)

Ghitulescu
21st July 2009, 12:25
mistifing professions is probably the reason for this thread (don't do it, Lucas is not a god)
I've seen "Funny face" on DVD with film grains and once more, recently, in 720p on ARTE HD (no grain at all). I assume they used similar quality prints, and denoising HDTV resolutions is relatively difficult (the grain should be of a similar size, only the pixels become smaller). So it is possible to degrain a print without loosing details or introducing artifacts in the image.
I'm not making anyone God, I do own professional apparatusses and I know what they can do. I was talking about a job "professionally done". Do you think they run Ice Age through CCE in 6 passes with 8Mbps over night? No, they run it through a real time encoder. The very same is also valid for SAT transmissions (their encoders are however even more complicated as they code several streams at once and distributing the bits also between the concurrent streams, all in real time).

*.mp4 guy
21st July 2009, 14:04
A lot of noise on dvd's, especially older ones, is not from the original film stock, its very common for a new transfer to have much less noise then an older one, just because better methods were used, perhaps using a less degraded filom print as the source. Also, "profesional quality" denoising is often worse then what can be done with avisynth, so no the profesionals have no idea what they are doing.

Ghitulescu
21st July 2009, 14:37
I'll stop the argument here, because my point was made: it is possible to remove the film grain without touching the image. I've seen it done.

I don't care how it was made, by whom and where. It was done, and by professionals, since ARTE is a broadcaster. I am surprised to see how good is ARTE team in doing such restorations, as most of their broadcasts are incredibly clean.

I am sure that most of you can use avisynth to get a even better quality than that of most older DVDs, but this is because you have what a pro sometimes doesn't: time. So they are glad to see that there is a "film grain supporting community" because that would simply mean they are no longer obliged to clean the movies.

smok3
21st July 2009, 17:53
I am surprised to see how good is ARTE team in doing such restorations, as most of their broadcasts are incredibly clean.
any samples? (before/after)

nobody said that grain is holy grail, so you missed your point (movies are cleaned to remove all kind of other problems as well, scratches, ect)

Do you think they run Ice Age through CCE in 6 passes with 8Mbps over night?
funny that you mention that, i just did a quick x264 crf 21 encode (that's single pass) from a DVD source and the quality is more than decent, nope most cartoons aren't that problematic to encode.

Ghitulescu
21st July 2009, 19:22
any samples? (before/after)

I don't have access to their videoteque, however, their movies are extraordinary worked out. You probably do not receive it, otherwise you would know by now what I'm talking about. I'll try to remember to capture a still next time Funny face will on air. Or any movie made before the WWI (that's 1914-1918) or WWII.

nobody said that grain is holy grail, so you missed your point (movies are cleaned to remove all kind of other problems as well, scratches, etc.)

As I said, I don't care what they did. Final point: video quality. Grain is a defect. Actually, the main issue here seems to be that some people think that film grain adds something to the "personality" while others disagree. I'm one of the other guys. I prefer to see the movie without grain, scratches, color defects etc. the whole bullshit studios try to sell as "film personality". Who wants genuine grain or scratches has the choice of a 16mm projector, but I'm afraid the choice of getting 16mm films is next to nothing, unless s/he buys him/herself a film printer and copy the DVDs. That one will also have the real feeling as s/he will hear also the shutter, like our great grandfathers did. A short wing piano would complete the studio.
funny that you mention that, i just did a quick x264 crf 21 encode (that's single pass) from a DVD source and the quality is more than decent, nope most cartoons aren't that problematic to encode.
I said it before, I don't either care what you or others do, chances are that I'll never buy from you/others. What I care is that the movies I buy to be of a good quality (no grain among other issues). However, if some community feels that grain bring "personality" to a movie, this the excuse that the studios need to do exactly what their music counterparts did: MP3 with a loss of dynamic and quality in the original mix, because why bother employing state-of-the-art mixers/digitizers/recorders if the audience don't give a damn shit, they simply like MP3 because you can pack 10 songs instead of one. In other words, they will bring new movies on the market that have no story but what nice 3D effects they have and what "kino feeling" they give with their grain. And they can buy another house on the beach on the money they just saved from cutting post-processing costs - the audience demands grain in the movies.
You want grain: nothing stops you adding it to a clean DVD/BD while recoding it into h.264 (x264 crf 21).

smok3
21st July 2009, 19:29
crapy mp3's are exactly what the public demanded, so you are right (loudness war has nothing to do with perfectly fine mp3 compression btw, but lets not dig into that...), actually they kill things during the so called 'mastering' process, so even if they get a correct mix is of no use....
http://en.wikipedia.org/wiki/Loudness_war (the way to stop that is to tell them that there is something called replaygain, and that their 'mixes' will get extra penalty in loudness if they overcompress)

another house on the beach on the money they just saved from cutting post-processing costs
they will bring new movies on the market that have no story but what nice 3D effects they have and what "kino feeling" they give with their grain

that is certainly a factor, why would one stare at the telecine scanner, when it could watch dolphins jumping around. :D (most video/film stuff is about the budget actually...).

You want grain: nothing stops you adding it to a clean DVD/BD while recoding it into h.264 (x264 crf 21).
i will not modify what others did (to that extent), but i could do that (without any hard feelings) on one of my own edits, yep. Actually i demand to see the movie exactly as it was intended by the director.

p.s. nope, i don't watch arte, but i do remember one of theirs excellent promo campains (maybe it was Promax/Amsterdam, not sure).

p.s.2.
I'll try to remember to capture a still next time Funny face will on air
better capture a short 30s video clip, stills don't tell much usually

p.s.3. btw, the reason why video is denoised/smoothed is usually not about the correctness of the procedure (and because they love their public), it is actually about stuffing more into their multiplexes (which are usually of the limited bandwidth).

Soulhunter
21st July 2009, 22:25
I don't see noise in real life. I see noise on TV and DVDs, but again not in real life. If one see grains while looking at a tree, a lake, a meadow, it should go to a doctor - has a visual problem.
He says everything is fine and its absolutely normal... :)

I'll stop the argument here, because my point was made: it is possible to remove the film grain without touching the image. I've seen it done.
Then YOU seem to have a visual problem, because this is more or less impossible!

Adub
22nd July 2009, 00:01
I agree. It is a mathematical impossibility to remove "every piece of grain and keep the details".

Ghitulescu
22nd July 2009, 07:35
I agree. It is a mathematical impossibility to remove "every piece of grain and keep the details".

Some 100 years ago was also mathematically impossible for an apparatus to fly if it was heavier than the air ;)
Some 150 years ago it was mathematically demonstrated that the passengers of a train would suffocated due to lack of air if the train travels faster than 37kmh (I think).
:D

Mug Funky
22nd July 2009, 08:08
okay, i'm still going to bite even though we're all being trolled.

for starters, the "professional" NR tools are at best equal to avisynth (we're talking $500K for this gear), and for the most part are the same as peachsmoother or avisynth's temporalsoften. the main difference is 10 bit or greater versus avisynth's 8 bits. and that's a very small difference that can be overcome. the DVNR 4x4 we've got here is basically a 5 RU peachsmoother. true, it can be controlled shot-by-shot, but i usually leave it either off or on it's lowest setting (modern film stocks don't need it anyway - more harm than good).

PFClean is apparently quite good, but i've only used it's de-spotting tools.

avisynth basically represents the state of the art in noise reduction. it's only problem is this 8-bit barrier that it really needs to get through (in my humble opinion). even then, 8 bits is fine on a graded image, or for use as an element in a composite image. back in the day, film-out recordings were done in 8 bits... and for the most part the grain hides it nicely :)

where's this notion of grain being a "defect" come from? it's inherent in the medium. that's like calling pixels a defect! the picture is made from the grains!

this "funny face" transfer you speak of. i'd like to see some a/b comparisons between the HD and SD transfers. colour muck aside, i reckon you'll see details missing that are present in the SD transfer, because old films are grainy as hell, even at slow speeds.

another explanation for film formats getting smaller and smaller is one of technology and economy. it is now possible to get less grain from 16mm, 500 ISO than you could from 35mm 50 ISO not 20 years ago. the current line of kodak vision3 films are amazing. filmmakers are routinely cutting 50D with 500T these days, and the results are practically seamless.

true, it'd be nice to shoot on IMAX for everything, but we're not all that rich.

one explanation for the increase in grain you may be noticing in feature films is simply the fashion of having very high contrast colour-grades. when that process went digital it became possible to do that kind of thing, and so everyone's doing it. these days film is also much sharper for the same reason - no blurring in intermediate films. sharper prints means more visible grain. also, DVDs and Blu-Ray discs are coming straight from a digital intermediate these days, so there's only one layer of film between the camera and your TV. this means it's as sharp as it is on the original camera negative. this means grain is going to show a hell of a lot more.

i'm not saying noise reduction is evil - i spend a lot of time trying to perfect. i am saying though that a person's default avisynth script should be empty. noise reducers only added if the encoded result is crap (with x264 being the way it is, is almost never the case).

Ghitulescu
22nd July 2009, 08:29
where's this notion of grain being a "defect" come from? it's inherent in the medium. that's like calling pixels a defect! the picture is made from the grains!

I know a image is made of pixels, even on analog media (coloured spots ;)). The grains appear however evident when their size depasses that of the smallest pictural unit on the displaying device. Id est one grain is spread over 2 or more pixels. The next image the same pixels encompass different parts of grains and so on. because the pixels are ALWYS in the same position, while the grain "moves" due to an inherent "missmatch" of their particle size relative to film speed travel (advance step - don't know this in English, it's the distance a film goes within one exposure).

this "funny face" transfer you speak of. i'd like to see some a/b comparisons between the HD and SD transfers. colour muck aside, i reckon you'll see details missing that are present in the SD transfer, because old films are grainy as hell, even at slow speeds.

I do not own the DVD and I have to wait for a rerun in 720p to have a clip (in August probably).

As I previously said, I don't care about the grains if they are not visible (they have a comparable size or less than that of a pixel). In this case they are a feature. But having "running" (random) grains on my screen, that's a defect. They do not move synchronously with the picture they depict - that's the defect.

Ghitulescu
24th July 2009, 08:39
I've checked again my Humax, luckily I've found several minutes of it. The file is too large and I cannot cut it yet, but I did a snapshot (in the attachments). There are small artefacts in this image, but the overall impression on a 50" FullHD plasma is good, considering the age of the movie.

Lyris
11th August 2009, 01:15
Personally, that looks very unappealing and artificial.

kieranrk
11th August 2009, 12:33
I have to agree with Lyris.

They look like clay figurines.

Ghitulescu
11th August 2009, 13:14
It's not the best shot but I chose it for it depicts dark zones, which are most prone to noise (both in digital and analog).

Because we discussed here the noise and the film grain and not the "clay aspect" of the characters, which is another story that cannot be linked to denoising/degraining.

kieranrk
11th August 2009, 17:28
Because we discussed here the noise and the film grain and not the "clay aspect" of the characters, which is another story that cannot be linked to denoising/degraining.

I always thought the clay look was because of DNR.
Can you explain further please?

Ghitulescu
11th August 2009, 18:32
I always thought the clay look was because of DNR.
Can you explain further please?

My mistake, I thought you referred to the colours, which indeed are a bit more saturated than one sees in a kino. I interpreted "clay" as shiny [painted] pottery (glazurated).

I chose this snapshot because it was a dark one (which poses problems both in analog and digital) and contained a lot of details (books, pages in the book, Audrey's dress). A denoising process would have had a lot to do! And the movie is really nice to be watched (the colours are a bit oversaturated, but I think that was the fashion back then in the '60ies).
-=-
To be clear once more: I'm not talking here about the grain that is part of the colour substrate of the film, I'm talking about the noise that people mistakenly confounds with the first one, but it's not despite it has the same source, just bigger. You'll see it sometimes in trailers, or used as a comparative example for image enhancement processes. Or in analog TV (who still has it) with a subpar reception. I'll try to find a snapshot (the day of the locust - DVD9) or this one (demetrius and the gladiators-DVD9).

-=-
There is also a reason why some artefacts are present in both images but not in the original ones: the .PNG image is compressed into a .JPEG one (for the last image, from 830kB to 40kB).