Question for Mattias (or any other compression guru)

[MovieStuff]

Here is the question:

At the same level of visual quality, would one hour of 8mm film telecined take up more room on a DVD as one hour of video?

Now, this may seem like a play on the old joke, "Which weighs more, a pound of lead or a pound of gold?" But think about it. Video doesn't have a constantly changing grain pattern like film. If you wanted to maintain all the original grain of the film from frame to frame, then the bit rate would need to be higher than for the same scene shot on video, wouldn't it? I mean, let's say that nothing is moving in the scene and the camera is locked off on a tripod. For all practical purposes, a freeze frame of the video would hold all the detail necessary endlessly since nothing is really chaning from video frame to video frame. But a film transfer would have ever changing detail because of the random grain pattern not present in straight video. So if the bit rate is too low, wouldn't you lose that random grain pattern and, therefore, some of the visual quality of the film?

Scratching my head over this one......

Roger

[Scotness]

I know if I compress one clip with the dvsd crappy dv codec it removes alot of the grain but if I compress the same clip with huffyuv it keeps the grain but the file sizes are 3-4 times bigger.

So my money is on the lead...er um I mean film

Scot

[Royalbox]

I don't feel qualified enough to reply compared to others here, but I have burnt quite a few DVDs. I know that "1 hour" was just an example, but if you only need 1 hour then you would be able to use a near-maximum constant bit rate of 8000 and get the best quality from the DVD possible anyway.

But, I think with variable bit rate, then the encoder would use far less bits on your video example than it would with a "noisy" or grainy film transfer. So the video encoding would end up smaller than the film encoding.

I've really let myself in for a bashing I can sense it!

[MovieStuff]

But, I think with variable bit rate, then the encoder would use far less bits on your video example than it would with a "noisy" or grainy film transfer. So the video encoding would end up smaller than the film encoding.

Well, that is my thought, as well. Obviously, if using constant bit rate, then both would be the same. But if using variable bit rate, the grain would certainly add some "activity" that would have to be accounted for and that seems like it would increase the file size and cut down on running time. To that end, I wonder if regular 8mm would take up more room than super 8 and super 8 would take up more room than 16mm, etc? Obviously, there reaches a point where the grain is insignificant, compared to the frame size.

Roger

[boohoo]

What compression software? Mac or PC?

[MovieStuff]

What compression software? Mac or PC?

Why would that make a difference? The issue is about whether the active grain of the film means a larger file size because preserving the grain from frame to frame would mean a higher bit rate. If I had footage of a white wall on video and a white wall on film, the film would have more activity in the frame, even though there was no "action" going on. Would there be a difference between compression on a Mac or a PC? I don't understand.

Roger

[boohoo]

Are you sure. I have a college friend right next to me who does mpeg compression and dvd authoring in an industry I prefer not to name. He deals with this everyday and he seems to say it does. I'll ask him to post when he gets home.

[wallnuts]

What compression software? Mac or PC?

Why would that make a difference? The issue is about whether the active grain of the film means a larger file size because preserving the grain from frame to frame would mean a higher bit rate. If I had footage of a white wall on video and a white wall on film, the film would have more activity in the frame, even though there was no "action" going on. Would there be a difference between compression on a Mac or a PC? I don't understand.

Roger

Weeell,

An old friend of mine who used Macs a lot had generally smaller file sizes than a PC because of the program language difference between a Mac and PC. But that comparision is really apples to oranges here.

I generally agree with you Roger, but what about this... what if the video was recorded analog and not digital. The noise generated during an analog video recording would have to account for something, wouldn't it?

Andy

[sunrise]

Roger,

you're quite right when talking film frame compared to MPEG compression for DVD, since MPEG compression works in a way that it registrates the changes in the image and not the full frame. If you have a clear blue sky at the top half and ocean in the bottom half, the MPEG encoding will spend most of it's power encoding the ocean which is constant moving.

Talking about the ever changing grain structure transfered to film would ideally for comparision mean that one grain equals one digital pixel. It does not, for reasons well know: Super carries more resolution than digital video and the grain structure in super8 is not nicely organized in rows and colums like digital video.

When doing a transfer of a blue sky with grain structure, one digital pixel would read from a few of the films grain (varies from stock to stock) and it would take the middle value of these grain to get it's colour information. So you see even though you see the grain on a projected image, making the resolution smaller would remove the grain and you will loose colour information.

I was discussing this at a party this weekend with a guy who had worked scanning stills and restoring them or transfering them to digital format. He was convinced that when a good quality scan came above 2000dpi, you would no longer scan the image, but scan the grain. A higher resolution image would be better, he said, for magazines, web etc. since it was closer to how the screen and the printer works.

I believe he might be right. But this discussion about film vs. digital that always goes on here is not taking this into consideration. It is not only about which media gives the best resolution but also what the transfer method is, and how good (or what happens) in the transfer and what the final format is. Film suffers in this way, since it's powers does not lie in digital projection; there is a strong decrease in quality along the process.

Hope that maked sense...

michael

[mattias]

it depends. film is smoother and has progressive frames, but it also has grain and image stability problems. i'd say there's no difference at these dvd bitrate levels. the relatively mild compression is designed to not have any problems with grain and such. the motion characteristics on a grander scale are much more important.

as for the mac/pc thing, any difference is obviously due to using different encoders. the same encoder on both platforms obviously produces the exact same result. and believing that macs and pc's would be differently suited to compress either film or video material is just plain silly.

/matt

[VideoFred]

About the grain....
Is this realy the film grain? According to Kodak, the grain size
of Super-8 film is between 0.1 and 3 micron! (1 micron=1/1000mm).
This is less than 1 pixel.

Suppose the grain is 2 micron. The Super-8 image is about 6x4mm.
Compared with digital this would be a resolution of 3000x2000 pixels.
Or is this a bad comparison? (I know the film grain is not equal)

I have the same 'grainy' effect with my
poor and simple 640x480 webcam, on dark scenes.
This could never be the film grain, on this resolution.

Is this not some kind of digital effect?
Or maybe it is indeed the film grain, clothing together...

Fred.

[mattias]

the grain we see in film is due to the grain structure, where grains group together differently depending on the image characteristics around them. this is why large areas of the same color look grainier for example -- at an edge the grain forms in lines rather than two dimensional structures. you're absolutely right that the individual grains are way too small to be seen on video though.

/matt

[unxetas]

About the grain....
Is this realy the film grain? According to Kodak, the grain size
of Super-8 film is between 0.1 and 3 micron! (1 micron=1/1000mm).
This is less than 1 pixel.

Suppose the grain is 2 micron. The Super-8 image is about 6x4mm.
Compared with digital this would be a resolution of 3000x2000 pixels.
Or is this a bad comparison? (I know the film grain is not equal)

I have the same 'grainy' effect with my
poor and simple 640x480 webcam, on dark scenes.
This could never be the film grain, on this resolution.

Is this not some kind of digital effect?
Or maybe it is indeed the film grain, clothing together...

Fred.

what you see on your webcam is noise.. when video cameras start getting low light levels, they sometimes amplify the signal to be able to get some image.. that also increases the dB levels or something like that..

[VideoFred]

What you see on your webcam is noise.. when video cameras start getting low light levels, they sometimes amplify the signal to be able to get some image.. that also increases the dB levels or something like that..

Maybe... But everything is set to 'manual', amplification at the lowest possible.. Adding more light from the spot does not help.. Setting the iris of the lens to a larger opening does help. This means, if the CCD gets more light, it's better. So, maybe you're right, and I can't control
whats happening behind the scene, deep inside the chips of the camera.

But what about the others with better camera's?
I think Matt is right here: grouping grains, depending on the colors of the scene.

Fred.

[boohoo]

I asked two questions. One about mac or PC. The other question was about SOFTWARE. There is a noticable differnce in file size and quality from software to software. $3000 mpeg compression software will make a big differnce when comapred to a $50. By naming his specific software, individuals may be able to give him spesifc instructions because no two softwares are alike. Thats when the mac or PC thing becomes important. please keep my posts in the context of which I stated