Noise in Digital - Grain in Film

[Angeliki Makraki]

Would someone please explain how grain in Film compares to Noise in Digital ?

Let's say I shoot a high- speed film for the effect and grain, is it equivalent to shooting

with an inferior digital camera for the noise ?

[David Mullen ASC]

Aesthetically, grain sort of randomly "swirls" with slight size variations whereas noise is more of a faster-moving "shimmer" or sparkle (it seems more rooted in position) -- it's difficult to describe using metaphors and similes. There is definitely something electronic/electrical about the look of noise (of course.)

 

Practically, it can sometimes be hard to tell the difference.

 

Grain is most visible in flat areas of middle tones, whereas noise seems more visible (to me) in the darker areas although I guess both are everywhere through the frame. Noise though can sometimes be more localized to a particular color area, although some colors in film too can be grainier than others.

[Chris Keth]

I think of it like this, perhaps it'll help:

 

Grain in film is like tile and glass bits in a mosaic. They are actually part of the image.

 

Noise is like gravel sprinkled on top of the image. They are not part of the image but rather little places where there is no image information on that frame.

[Michael Nash]

I think of it like this, perhaps it'll help:

 

Grain in film is like tile and glass bits in a mosaic. They are actually part of the image.

 

Noise is like gravel sprinkled on top of the image. They are not part of the image but rather little places where there is no image information on that frame.

 

That's a good analogy. The term "noise" after all means extraneous information that's not the signal. With film, the grain makes the signal, or is the signal.

 

I'll second that noise and grain can often appear similar, and can be used creatively to similar effect (who hasn't turned up the gain on a video camera to make a more "gritty" or "grainy" image?). It's also true the bit about grain showing first in midtones, and noise from gain showing up in the blacks first.

 

Another thing about noise is that it can be made up of random colors, sometimes looking like colored speckles added incongruously to the image. Film grain usually shows as black until it appears in very dark areas (where you can see more white speckles), and since it's integral to the image it just ends up looking like a rougher textured image, at least most of the time.

[K Borowski]

If I only had account space I could upload some comparative pictures depicting film grain versus digital noise. THere's a huge difference on the big screen, but on TV (especially SD) not so much, and also with multiple generations of copying it gets hard to tell the difference sometimes. If you copy digital enough times on film, the digital noise actually starts to become indistinguishable from underexposed negative film printed thorugh several generations, at least to my eye.

 

To me the biggest problem with digital continues to be its poor latitude and terrible rendition of flesh tone. I honestly do not understand how the chip makers didn't get fleshtone rigth before coming out with digital cameras, be they still or movie cameras. Back when the big K was working on Kodachrome (actually two violinists invented it, contracted and funded by Kodak), that was of prime importance to them. In fact film is still optimized to deliver good skin tones above other things for the most part, the only exceptions being certain slide films that are designed for photographing plants and scenery.

 

THere's something about digital skintones that is unnatuarally color-biased and "unreal" to my eyes, like a pink or orange or purple cast to people's faces. Sometimes it can be very pronounced. I've heard that overexposing digital can make this problem worse, but even properly exposed digital I have seen tends to show this artifact.

 

I think both grain and noise are undesirable elements and I try to minimize them in my work. At least grain is random though, instead of in pronounced block-like formations of random color.

[Angeliki Makraki]

I like the gritty feel of grain on film but not so sure about digital noise yet.

Thank you everyone for all your responses.

[Michael Nash]

THere's something about digital skintones that is unnatuarally color-biased and "unreal" to my eyes, like a pink or orange or purple cast to people's faces. Sometimes it can be very pronounced. I've heard that overexposing digital can make this problem worse, but even properly exposed digital I have seen tends to show this artifact.

 

I think what you're responding to is the more limited color depth of digital, compared to film. It's not that the cameras get the color wrong, it's that it doesn't contain as much subtle variation. The end result can be a little artificial or "plastic" looking. Overexposing video cameras usually renders skin tones with an excess of saturation relative to the luminance, at least before the knee kicks in (where you can sometimes control the amount of saturation). Some of the "purple" artifacts can come as a result of compression or heavy color correction; I usually only see it with bad MPEG compression, not so much in a camera original.

 

You touch on a good point about digital noise not holding up well on a big screen -- the size of camera-generated noise is relative to the pixel count of the camera. HD noise is smaller or "tighter" than standard def noise, similar to the way 35mm grain looks tighter than 16mm grain. The more you magnify either noise or grain, the more pronounced it looks; but when you shrink it down to standard DVD size you'll find that noise from an F-900 is less objectionable and more film-like than what you get from a DVX-100.

 

I don't know that I agree that digital noise is always "block-like." Usually anything "blocky" is an artifact of compression, like color banding and pixelated fades. Electronic or digital noise (created by gain-boosting or heavy in-camera processing) usually looks finer and more random.

 

But digital noise can come from a variety of sources at any stage of the process, including compression artifacts, and can take on different characteristics. Film grain can only come from one source -- the film grain itself (or perhaps silver in the case of silver-retention). Even if you dupe elements and do optical blowups, the grain is always film-originated.

[Phil Rhodes]

I think the thing that most differentiates noise from grain is that noise tends to be more chromatic. Grain is effectively an optical aberration, a refraction or shadow of the edge of a dye cloud. While this can subtract one of the RGB layers from a point in space and thereby alter its colour, the effect of grain is fundamentally in value, not chromacity.

 

Noise, especially if it is induced in YUV and then translated to RGB (as is very common in practical systems) can cause quite severe chromacity deviations.

 

Phil

[Jaan Shenberger]

I think the thing that most differentiates noise from grain is that noise tends to be more chromatic. Grain is effectively an optical aberration, a refraction or shadow of the edge of a dye cloud. While this can subtract one of the RGB layers from a point in space and thereby alter its colour, the effect of grain is fundamentally in value, not chromacity.

 

Noise, especially if it is induced in YUV and then translated to RGB (as is very common in practical systems) can cause quite severe chromacity deviations.

 

Phil

 

very good point by phil. video noise will hinder your ability to heavily color grade footage that has noise in any greyish/unsaturated areas because there is a random changing of hue and saturation on a per pixel basis amongst the noise and it will become pronounced if you push the color characteristics too far. hope this helps.

Edited April 4, 2007 by Jaan Shenberger

[K Borowski]

I think what you're responding to is the more limited color depth of digital, compared to film. It's not that the cameras get the color wrong, it's that it doesn't contain as much subtle variation. The end result can be a little artificial or "plastic" looking. Overexposing video cameras usually renders skin tones with an excess of saturation relative to the luminance, at least before the knee kicks in (where you can sometimes control the amount of saturation). Some of the "purple" artifacts can come as a result of compression or heavy color correction; I usually only see it with bad MPEG compression, not so much in a camera original.

 

Thanks Mike, you put it into words much more succinctly than I did. That is exactly what I meant, although even digital backs or raw files from full-frame 35mm DSLRs "trouble" my eyes, not just HD video that's heavily compressed. There's something inately part of the look of the digital process that bothers me, something I can't get used to.