h a l f b a k e r yFewer ducks than estimates indicate.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Please log in.
Before you can vote, you need to register.
Please log in or create an account.
|
One of the common goals of digital film restoration is to rid movies of the visual defects associated with dust, scratches, mildew, splices, and other such ills. A major element of common methods is to compare consecutive frames of film to each other and use various methods to determine which features
on a frame are real and which ones are defects (that should be painted out using adjoining frames).
Unfortunately, computers aren't very good at determining such things, and consequently can tend to remove or blur things that are supposed to appear on the film.
I would suggest that movies be scanned using a device which would photograph each frame from the emulsion side and from the base side, using a very large-aperture short-focal-length lens, taking pictures at several distances and with front-lighting (which, in the absense of defects, should show mostly blackness), side-lighting, and backlighting (which should show the proper picture).
If a movie is completely free of dirt, scratches, warping, and other defects, the backlit photos of the both sides, taken at the nominal focal distance, should match and should be quite clear. Any dust or scratches, however, will almost certainly appear in the front-lit and/or side-lit shots.
Thus, a system which tries to digitally clean up the film will be able to identify what 'spots' are a result of actual photographed material (and should be left alone) and what spots are a result of dust, scratches, or other defects and should thus be painted out.
(?) Digital ICE
http://www.asf.com/...lmICEOverview.shtml
Digital photo restoration. [trekbody, Feb 22 2005]
[link]
|
| |
What's "holographic" about this? |
|
| |
It may be far less costly (and possibly equally good) to use diffuse illumination from the front and image capture from the back. Diffuse frontal illumination tends to cancel out the effect of scratches, dirt, etc. |
|
| |
This doesn't sound at all holographic. This is also pretty close to one of the techniques used in high-end Nikon negative scanners. |
|
| |
Not truly holographic, but much closer to a 3d capture than a conventional scanner (due to the use of multiple scans from different directions and distances, and using different lighting) so as to be able to determine which defects and features are in the same plane as the emulsion and which ones are not. |
|
| |
Although I didn't go into particular detail, the use of wide-aperture short-focal-length scanning at different distances could probably captures some information about water spots and other defects which could not be captured by any other means. |
|
| |
Or use a wet gate telecine. |
|
| |
A wet-gate telecine will help a lot, but it's not a cure-all. Having a means of explicitly detecting defects would help with efforts to remove them. |
|
| |
BTW, I would think that wet-gate projection and telecine would be quite simple if the film were horizontal. For projection, just use a mirror to put the film image right-way-up on the screen. |
|
| |
Also, bristols, I'm curious: will wetgate telecine help with film where the emulsion has become crazed (a 'spiderweb' effect of transparent cracks)? |
|
| |
Wet gate can go a long way but it isn't perfect. I don't know about crazing but imagine that it would provided the emulsion itself isn't disturbed. It can't create image where there isn't any. |
|
| |
Not sure anybody is doing anything recent with wet gate technology. |
|
| |
My suspicion is that the crazing is a result of the emulsion BEING disturbed, and I wouldn't expect wetgate to help much. But something like the multi-light scanner I described could probably create a very accurate picture of the crazing itself. Such a picture could be used to know what parts of the picture needed to be 'filled in' using spacially or temporally adjacent material. |
|
| |
Isn't this baked already with Digital ICE for scanners? (See link). |
|
| |
Well i hope you get this all figured out. I'm theorizing there is 3D information trapped in camera negative as well that can be pulled out. |
|
| |
It would appear that DigitalICE encompasses certain features of the idea (thanks for the link), including the concept of deriving an error mask. Doing things like shooting both sides of the film under different lighting conditions goes far beyond Digital ICE, however. |
|
| |
For example, I would expect that with the right lighting and scanning equipment, it should be possible to read the emulsion image off a stretch of film whose emulsion or base has a completely opaque defect. The image data thus read would probably not have the same quality as would conventionally-scanned data from undamaged film, but could nonetheless be useful in a restoration effort. |
|
| |
mensmaximus: A conventional camera shooting on conventional film will not record anything particularly useful in the way of '3d' information from the subject, because the film is much flatter than the subject. |
|
| |
If one were to use an extremely-thick-emulsion process, 3d information could be recorded though reading it out would be rather difficult. |
|
| |
[supercat], I have hand-processed motion picture steady tests at home at one time and saw how the anti-halation backing came off in my hands after the rinse. I know they had a secondary reflected image,(spread), problem previous to this backing application method, coming from the base. Not much room for 3D theory there but I'm reasoning that more than one image is being focused by the lens, (varying from good to bad lens design) and that varying methods of scanning with acute spectral analysis of some sort may extract enough information that can be assembled to form the second image of a stereoscopic pair. Any halation there now may be an unfocused additional image of some sort. I have another theory of pulling 3D info that I will post soon. |
|
| |
And why can't you build some huge rotoscope in an airplane hanger and hire a few artists to paint in a few missing details. How much money can be spent per frame to save a day's shooting? |
|
| |
//Not much room for 3D theory there but I'm reasoning that more than one image is being focused by the lens, (varying from good to bad lens design) and that varying methods of scanning with acute spectral analysis of some sort may extract enough information that can be assembled to form the second image of a stereoscopic pair.// |
|
| |
If a lens has focal length f, a perfect image of an object that's distance d1 from the lens will produce a single image of the object at distance d2 such that 1/d1 + 1/d2 = 1/f. |
|
| |
In practice, most of the time an object won't be at absolutely the perfect distance, so the result will be a blurry image. |
|
| |
If the emusion is thick, an image which is perfectly focused on the front would be blurry on the back, or vice versa. |
|
| |
One could perhaps try to gather some 3d information about the subject, but it would be about as useful as a stereoscopic camera with two lenses 1cm apart. |
|
| |
//And why can't you build some huge rotoscope in an airplane hanger and hire a few artists to paint in a few missing details. How much money can be spent per frame to save a day's shooting?// |
|
| |
The goal isn't to salvage imperfect film that was just shot. The goal is to salvage film which cannot very will be re-created (say, Melies' The Troublesome Heads of 1898). |
|
| |
Don't know that I've ever actually seen a nitrate film, but I've heard it said that the picture quality is better than acetate. Scanning films with an apparatus such as this might reveal whether that was true or not. |
|
| |
Very neat. I just hope it's done soon, though-- nitrate film disintegrates (literally) after a hundred years or so. |
|
| |
//Very neat. I just hope it's done soon, though-- nitrate film disintegrates (literally) after a hundred years or so.// |
|
| |
So does acetate, and the time may be well under 100 years if storage conditions are not good. The big problem with nitrate film isn't just that it disintegrates, but that the products of such disintegration are extremely flammable and may sometimes self-ignite. |
|
| |
Take another tangent; good idea, wrong application in my opinion... |
|
| |
Decomposing nitrate, diacetate and acetate film sometimes develops dimensional instability in the form of blisters, buckles or waviness that is difficult to "smooth out" optically due to varying image geometry. This scanner could contribute to a smoother playback by using some form of mesh deformation algorithm to "de-warp" the image. Using a "normal" sample of a bit of film in the reel that is fairly flat and undeformed, you could map the deformed frames to this "ideal" mesh and hopefully remove some of the geometric distortion. |
|
| |
Of course, I am speaking of terminally decomposing film; film that is often written off as being unrecoverable or beyond economical recovery efforts. |
|
| |
While the economic viability (i.e., profitability) of developing such a tool would be next to nothing, it would be a real boon to film archivists. |
|
| |
In fact, I was searching the internet for a pre-fab solution to this very problem when I ran across this site! |
|
| |
I'm somewhat curious about film restoration; how often will a film have usable images and yet be too brittle and deformed to scan using conventional means? If a film became impossibly brittle, and yet the emulsion was still stuck soundly to the base, I would think it might be possible to carefully use a knife to cut off strips of 2-6 frames at a time and scan them; the holographic scanner idea could help greatly if the strips didn't lie flat (trying to force the strips to lie flat could easily destroy the images thereon). I would think some restorationists would cringe at the idea of chopping a roll of rare movie film into little pieces, but using such an approach one could avoid having to flex the original material at all. |
|
| |
Of course, that leaves open the question: if movie film has degraded to that point, is the emulsion likely to remain attached? |
|
| |
It is very hard to generalize how film degrades, but, yes, you can have a film so brittle it cannot be unrolled from the reel without it snapping AND still have very viable images on the emulsion. |
|
| |
( too bad you cannot post images here or I would show you an example ) |
|
| |
When the film HAS reached this point, archivists are all for whatever is required to copy the film if it is unique and in active decomposition. |
|
| |
The film is going away anyway, so either do nothing and loose everything OR try whatever you can and retain something. |
|
| |
Your idea to do it in strips is not unique; it is being done in several labs I know of -- they simply wind the film out on to a backlit table and use a digital camera to take frame grabs of the images for later digital processing. |
|
| |
Not perfect, but it DOES save something and can result in some decent images. |
|
| |
You would be amazed at the physical condition a roll of film can degrade to and still retain enough information to make it worthwhile to copy. |
|
| |
If it is the ONLY known copy in the World, then you try real hard to recover it. |
|
| |