True Resolution

[Landon D. Parks]

Does any one know what the true resolution of Super-8 film is? Im guessing somthing equal to 300,000 - 500,000 pixels... But I dont really know. :rolleyes:

[David Mullen ASC]

Again, film does not have pixels. The real question is what minimum pixel resolution do you need to scan film in order to reproduce all the detail on the film. I mean, you could scan Super-8 at 60K but I'm sure once you get above a certain point, you're not gaining anything.

 

The other problem is that while you could take the resolution data for the Kodachrome 40 stock (probably one of the sharper stocks for Super-8, that and the new Vision-2 negative stocks) it doesn't take into account the softening due to registration problems and lens optics.

 

So while people like to say that Super-8 actually has the resolution of Hi-Def because of the data from looking at stocks like K40, the practical reality is that it never looks like it does. And the grain is large enough to make you think the image has less resolution.

 

Anyway, practically-speaking, I'd say a good scan at PAL or HD resolution would adequately resolve and reproduce all the information on the Super-8 frame.

 

But comparing Super-8 to something shot in 24P NTSC or 25P PAL is sort of an apples & oranges thing because the image characteristics are so different. I'd say that resolution would be similar to something shot on, say, the DVX100, but that grain and any problems with optics, registration, and the quality of the transfer to video can greatly reduce perceived sharpness, allowing the DVX100 image to be enlarged more successfully.

[Landon D. Parks]

I see.. not that I have any plans to shoot it. I have heard a lot about super 8... and how its a sin to shoot it (mostly from books).

[Phil Rhodes]

Hi,

 

A sin? Why is it a sin? Why does it matter enough to anybody for them to care? I think super-8 is quite good fun.

 

Phil

[Landon D. Parks]

I mean... "I" dont consider it a sin. I Dont consider somthing like that unless I have used somthing..

 

I have just noticed in these Filmmaking books when they interview people... they always say stuff like that.

 

But I dont know....

[John Pytlak RIP]

Super-8 is very useful in providing an alternative "look". As noted before, Super-16mm has the advantage of increased image area, greater stock selection, and more widely available processing and transfer facilities.

[Marty Hamrick]

The other problem is that while you could take the resolution data for the Kodachrome 40 stock (probably one of the sharper stocks for Super-8, that and the new Vision-2 negative stocks) it doesn't take into account the softening due to registration problems and lens optics.

How much difference would there be if one is shooting with a top quality camera like a Nizo,Nikon R-10,Beaulieu or Canon 1014?Would that not negate some of the problems associated with poor optics?

Would shooting double super 8 with a real pressure plate negate the poor registration problems?

Combining those factors,if you shot DS-8 with really good glass would that get the most res possible with the super 8 film frame?

Marty

[John Pytlak RIP]

Film resolution is directly proportional to image area. So there is no way a Super-8 image will ever be as sharp or have as low a graininess as a larger format image, all else being equal. With a smaller format, "good glass" is all the more important, and any issues with camera/telecine registration are magnified more.

[Elhanan Matos]

Here you go, this will teach you a bit more about resolution.

 

http://www.normankoren.com/Tutorials/MTF.html

[Matt Sandstrom]

*delurk*

 

i shot a test chart with k40 in a canon 814e with the zoom on medium and stopped down to 2.8 or so a few years ago, and projected it with a high end projector. it indicated a resolution of the equivalent of around 800x600 pixels, which seems to match what david mullen is saying about it being the ballpark of pal and hd. this was for the moving image though. the grain makes the still frames less sharp.

 

/matt

--

http://www.mattias.nu/

[Carl Looper]

The limits on resolution for Super 8 are the lens used (including aperture) and the size and distribution of the silver particles (monochromatic stock) or dye clouds (colour stock). New filmstocks from Kodak for Super 8 have a very fine grain (less noise) than older stocks.

 

I've done a digital scan of Tri-X Super 8 (3000 pixles wide) using a microscope objective and the result is a lot better than the same scanned at less. But the grain is more pronounced. This is not, in itself, actually that offensive. But it does demonstrate how the resolution is limited. When you look closely you can see what should be a straight line start to break up into noise.

 

However - using temporal integration techniques one can, in theory (and also in practice), reconstruct a higher definition signal from the Super 8 signal (increase resolution). The technique is called "super-resolution" and there are numerous papers on the web. Note that this is not just interpolation of the signal up to a "higher" definition signal but reconstruction of an actual higher definition signal already embedded in the Super 8 film (across a number of frames).

 

One can think of it this way: The Super 8 image contains a compressed signal. The super-resolution algorithms are attempts to decompress this signal.

 

To increase the resolution of Super 8 to that of raw 16mm would require a 2X super-resolution process (4 frame integration) because 16mm has about 4 times the area of Super 8. For best results the Super 8 film should be shot with a high quality lens and scanned at the highest definition possible (eg. 2K wide scan or better). And of course you'll need a good algorithm. This remains an active area of research.

 

Note that limits to super-resolution for film are less limited than the same proceedure for video/digital due to the random distribution of particles/dye clouds in film.

[David Mullen ASC]

Wow, a reply to a 6-year-old thread... makes me wonder what happened to Landon, he was so young. But that was six years ago, he must be in his mid-20's by now.

[Brad Grimmett]

You crack me up David.

Funny, I saw the post with his name on it and thought, "Wow, he's back." I guess you did the same.

[Carl Looper]

I didn't realise the thread was that old when I responded. Of course the information I posted is for anyone doing a search on Super 8 resolution.

 

I shot endless feet of Super 8 in my youth (the late 70s) and have recently picked it up again having worked in the digital domain, as a software developer, for the better half of my adult life. What has brought me back to it is the new filmstocks Kodak has released over the past few years, and the improved potential of Super 8 in the digital age.

 

Carl

[Friedemann Wachsmuth]

However - using temporal integration techniques one can, in theory (and also in practice), reconstruct a higher definition signal from the Super 8 signal (increase resolution).

 

Thats great insight, Carl -- and a very good read!

I recently wanted to "try out whats possible" and got a roll of Super-8 Ektachrome 100D (exposed in my Nikon R10) scanned by some guy in Holland who has built the IMHO best frame scanner ever, at least in the 2K space.

I then used avisynth (and an altered version of Freddy van der Putten's script) to "increase resolution" (among other things). During this I also scaled the scans down to 720p -- this is s single frame result:

 

Before avisynth optimization, the (almost) same frame looked like this. Its indeed quite impressive what software can do, especially free software in this case. And its even more impressive how much data such a little spool actually contains (and how failure tolerant its "encoding" is!)

Edited October 13, 2010 by Friedemann Wachsmuth

[Carl Looper]

Hi Friedemann,

 

that is such a very satisfying image.

 

Carl

[Will Montgomery]

Wow. I will have to investigate avisynth. Truly an amazing difference.

[Friedemann Wachsmuth]

Wow. I will have to investigate avisynth. Truly an amazing difference.

 

Good starting point: http://forum.doom9.org/showthread.php?t=144271

It does require quite some fiddling, but its fun. And free.

[Carl Looper]

Does any one know what the true resolution of Super-8 film is? Im guessing somthing equal to 300,000 - 500,000 pixels... But I dont really know. :rolleyes:

 

The resolution of Super8 is a function of a number of factors, each of which plays a role, including lens and filmstock.

 

A very good paper on this remains:

 

http://cool.conservation-us.org/coolaic/sg/emg/library/pdf/vitale/2006-03-vitale-filmgrain_resolution.pdf

 

It is getting a little dated now but still remains one of the best available papers on film and resolution. Of interest is the author's concluding remark:

 

"Software for Diminishing Film Grain

I have yet to find any software that reduced grain without affecting the image in the size

domain of the grain. I’m still looking for better grain removal software, but I’m less hopeful

because the problem is not well understood by the software engineers." - Tim Vitale

 

Software developers (such as myself) are now quite familiar with the subject - if only because of excellent papers such as this one. Grain, as the author clearly argues, is a "perception" rather than a physical quantity. If you observe any filmstock closely (eg. under a microscope) you will see more grain than if you observed the same filmstock at a distance (eg. projected on a screen at a small size).

 

Grain is related to, but is not the same thing as the silver particles (or dye clouds) in the emulsion. It is an observer determined quantity.

 

Kodak specifications of grain (granularity) standardise the observer (an aperture of fixed standard width etc). But all this allows is for comparison of filmstocks relative to each other, rather than how grainy a filmstock appears to an arbitrary observer.

 

The "de-graining" software about which the author complained were solutions that performed spatial integration - which is effectively the same as making an image smaller relative to an observer. If you increase the size of an image (relative to an observer) you increase the grain.

 

As mentioned in earlier posts the correct solution is to use temporal rather than spatial integration. You keep the image the same size, relative to an observer, but integrate grain along the time axis.

 

The more you can decrease the grain (along the time axis) the more you can increase the resolution for an image of the same size (relative to an observer).

 

The grain is functionally related to the size of the silver particles (or dye clouds) but is, importantly, independant of the signal recorded by the filmstock. It is independant of the signal's resolution. The grain, from one frame to the next (time axis) is statistically self cancelling, while the signal for a typical shot is not (because a shot typically maintains a correlated transformed version of itself from one frame to the next).

 

A human observer, watching a grainy film, can often sense, despite the grain, a high definition signal lurking beyond the grain. They are not imagining that signal. They are subconsiously performing temporal integration of the signal.

 

The resolution of Super 8 (or any other filmstock for that matter) is therefore a function of how much you can succesfully separate the signal from the film grain - whether in software or your own perceptual system (eye/brain)

 

Carl

Edited October 19, 2010 by Carl Looper

[Will Montgomery]

Good starting point: http://forum.doom9.org/showthread.php?t=144271

It does require quite some fiddling, but its fun. And free.

Hmmm... needs a PC not a Mac...

[Dom Jaeger]

This is a very interesting topic, thanks for the detailed explanations Carl.

 

What seems really impressive is that the enhanced film signal retains its organic look.

 

I look forward to more info on this.

[Elliott Landy]

The resolution of Super8 is a function of a number of factors, each of which plays a role, including lens and filmstock.

 

A very good paper on this remains:

 

http://cool.conservation-us.org/coolaic/sg/emg/library/pdf/vitale/2006-03-vitale-filmgrain_resolution.pdf

 

It is getting a little dated now but still remains one of the best available papers on film and resolution. Of interest is the author's concluding remark:

 

"Software for Diminishing Film Grain

I have yet to find any software that reduced grain without affecting the image in the size

domain of the grain. I’m still looking for better grain removal software, but I’m less hopeful

because the problem is not well understood by the software engineers." - Tim Vitale

 

Software developers (such as myself) are now quite familiar with the subject - if only because of excellent papers such as this one. Grain, as the author clearly argues, is a "perception" rather than a physical quantity. If you observe any filmstock closely (eg. under a microscope) you will see more grain than if you observed the same filmstock at a distance (eg. projected on a screen at a small size).

 

Grain is related to, but is not the same thing as the silver particles (or dye clouds) in the emulsion. It is an observer determined quantity.

 

Kodak specifications of grain (granularity) standardise the observer (an aperture of fixed standard width etc). But all this allows is for comparison of filmstocks relative to each other, rather than how grainy a filmstock appears to an arbitrary observer.

 

The "de-graining" software about which the author complained were solutions that performed spatial integration - which is effectively the same as making an image smaller relative to an observer. If you increase the size of an image (relative to an observer) you increase the grain.

 

As mentioned in earlier posts the correct solution is to use temporal rather than spatial integration. You keep the image the same size, relative to an observer, but integrate grain along the time axis.

 

The more you can decrease the grain (along the time axis) the more you can increase the resolution for an image of the same size (relative to an observer).

 

The grain is functionally related to the size of the silver particles (or dye clouds) but is, importantly, independant of the signal recorded by the filmstock. It is independant of the signal's resolution. The grain, from one frame to the next (time axis) is statistically self cancelling, while the signal for a typical shot is not (because a shot typically maintains a correlated transformed version of itself from one frame to the next).

 

A human observer, watching a grainy film, can often sense, despite the grain, a high definition signal lurking beyond the grain. They are not imagining that signal. They are subconsiously performing temporal integration of the signal.

 

The resolution of Super 8 (or any other filmstock for that matter) is therefore a function of how much you can succesfully separate the signal from the film grain - whether in software or your own perceptual system (eye/brain)

 

Carl

 

I do high resolution scans of my 35mm color slide film. I have made them at 4,000 DPI with Nikon scanner, 5400DPI with Minolta scanner, and currently use an Imacon 848 scanner at 8000 DPI. At least four years ago I tested every available Grain Reducing software for MAC. I found that Neat Image was head and shoulders better than others. http://www.neatimage.com/ It is also available for Windows. I have not tested since then and they have some new versions which I may not have upgraded to, but I am still happy whenever I need to reduce grain in an image. I am planning to transfer a lot of Super 8 film soon but am not sure if this app is applicable to moving pictures.

[Elliott Landy]

The limits on resolution for Super 8 are the lens used (including aperture) and the size and distribution of the silver particles (monochromatic stock) or dye clouds (colour stock). New filmstocks from Kodak for Super 8 have a very fine grain (less noise) than older stocks.

 

I've done a digital scan of Tri-X Super 8 (3000 pixles wide) using a microscope objective and the result is a lot better than the same scanned at less. But the grain is more pronounced. This is not, in itself, actually that offensive. But it does demonstrate how the resolution is limited. When you look closely you can see what should be a straight line start to break up into noise.

 

However - using temporal integration techniques one can, in theory (and also in practice), reconstruct a higher definition signal from the Super 8 signal (increase resolution). The technique is called "super-resolution" and there are numerous papers on the web. Note that this is not just interpolation of the signal up to a "higher" definition signal but reconstruction of an actual higher definition signal already embedded in the Super 8 film (across a number of frames).

 

One can think of it this way: The Super 8 image contains a compressed signal. The super-resolution algorithms are attempts to decompress this signal.

 

To increase the resolution of Super 8 to that of raw 16mm would require a 2X super-resolution process (4 frame integration) because 16mm has about 4 times the area of Super 8. For best results the Super 8 film should be shot with a high quality lens and scanned at the highest definition possible (eg. 2K wide scan or better). And of course you'll need a good algorithm. This remains an active area of research.

 

Note that limits to super-resolution for film are less limited than the same proceedure for video/digital due to the random distribution of particles/dye clouds in film.

 

So what is the best transfer house to use to get the best quality? I have a lot of Kodachrome. I want to use it for large screen projection so I want the highest resolution. I am planning to transfer it to HD 1080P with Pillars. I figure this will give me the best quality I can get out of the Super 8 Kodashrome stock. Is this correct? Or should I transfer to an even higher resolution than that? If it will improve it, I would do the higher res in expectation of future improvements in projection resolution just as HD has improved over SD. Any recommendations of specific labs would be appreciated. I was looking for an HD wet gate transfer but have not found it after extensive queries. Do you know one?

[Will Montgomery]

So what is the best transfer house to use to get the best quality?

For the highest resolution you would want a 2K scan from a Spirit at 2048 x 1556 pixels. Then you don't worry about pillar boxes because you have the full frame at 2048 wide.

 

This is really high resolution for Super 8 and there would be endless debate over if it was worth the cost, but since you asked, there it is.

 

Cinelicious would do a great job for you not only technically with their equipment, but artistically in their coloring.

 

Pro8mm should be able to do it with their Millennium but I don't see 2K scans offered there; I'm sure they'd find a way to take your money.

 

I'm sure there are some other companies that would offer 2K Super 8 scans as well.

[Chris Burke]

spectra film and video will have there spirit, which will have much better optics allowing it to scan the entire frame at full 2k resolution, online this month. Until the next thing comes along, it will

probably be the best for Super 8