Is it possible to hide a digital adjustment?

[Eigil Skovgaard]

Hi,
I have a hypothetical question:
Is it possible to scan a color transparency film (yes!), then make digital corrections (yes!) AND THEN copy or expose the result back to the same type of film without leaving ANY traces of the digital corrective process?
Best regards,
Eigil Skovgaard

[Repairman]

Re: Is it possible to hide a digital adjustment?

You could certainly output a second 'original' transparency that would pass close inspection (that's how the process worked prior to outputting cd's, electronic transfer etc) but it depends how deep you want to dig. If you overcome the easy stuff like stepping and banding you may well see, in magnification, cloning marks or 'smeared' pixels. A posterisation map can also reveal where retouching was carried out. So, good enough?. Yes. Enough to fool the CSI guy's - I have my doubts.

[Eigil Skovgaard]

Re: Is it possible to hide a digital adjustment?

Thanks for your input, Repairman.
Would it really be that hard?
I imagine a process where the "doctored" digital image is reproduced on a large screen of highest resolution and re-photographed onto the new chrome-film from a distance and/or through filters that would "drown" all patterns from the digital process (down to the last pixel); and no digging - even the deepest - would find anomalies in the film material?
I don't know anything about this, so I am just fishing (and not from CSI).
If it can be done, one might expect an exclusive marked for such skills and consequently an expertise. There is always one guy who can.
If anybody from this forum hear of such skills (companies, websites, illusionists etc.), please let me know. Not that I have anything to be faked, but the problem interests me. It would render analogue film-materials as useless for critical documenting as digital photography.
Best regards,
Eigil Skovgaard

[Repairman]

Re: Is it possible to hide a digital adjustment?

Even with conventional film, the inherent film grain is visible way before any pixels would show themselves. I'm assuming input/output from high end devices here.
From an input standpoint, I am not sure a digital photograph would yield more detail information than a quality 10x8 transparency would but maybe someone else can help us with that! The pixel resolution from the trannie would be set by the scanner operator who could theoretically scan just a portion of the film in high rez whereas a similar crop of a digital photo and output to 10x8 may well be softened by interpolation.
The ability of the retoucher would also be a major factor in how obvious the edits would show up.

[klev]

Re: Is it possible to hide a digital adjustment?

I'll give this one a try. First of all your idea of a high resolution screen is complete crap. Now you have two film grain patterns, and you have been caught. Film recording relied on a high resolution recording film. Recording film grain is so fine that it won't show up anywhere else, so it's not an issue. You don't want that. You want something that looks like the film base you originally used.

Here's how you get that. You need a scanner that can effectively give you a density map. You need to be able to replace a set of densities relative to the film base on one transparency to another. It's better if you do this with an entire film strip to make it even less obvious. Do a practice run to make sure you can achieve this.

You need a way to degrain the result before you output to new film. Otherwise you may have some influence from the original grain pattern, even though it was at a similar physical scale to this one. There has been a lot of research into degraining methods, but it will probably be something based on fourier analysis, wavelets, markov fields, and whatever else I think makes interesting reading material.

By the way, if you ever want to troll your guests, put a book on differential geometry on your coffee table. Pick one with a misleading title so they don't avoid it, and make sure it doesn't contain the word topology (scares them off too quickly) .

Anyway so you need a way to degrain the result and output to the same film type with consistent results. You also need tools that are suitable for this. Photoshop's standard profiles probably aren't good enough. The tools are another issue. You need something better than a circular brush. It has to preserve tangent space in most cases, yet not follow an easily recognizable pattern. It shouldn't violate sharp boundaries, because that would be obvious.

Now photoshop or whatever other application also needs to be able to read from and write to this format based on differencing. You probably won't find something that works off the shelf, so now you need a partner in crime. I recommend checking TOR. You'll want to generate a set of RSA keys to enable secure communication, and you may want to pick someone from a different country.

Okay I'm done being silly, but there are a lot more factors than you seem to realize. I just had way too much fun with that.

[Repairman]

Re: Is it possible to hide a digital adjustment?

So Klev, is that a yes or a no to the original question!?

[Eigil Skovgaard]

Re: Is it possible to hide a digital adjustment?

Apparently this takes further reading ... and I have still not heard the final verdict (much like the impatient customer with the Cadillac Man: stop talking about engine and transmission and tell me if the f.. car can run). Can this be done? Has it ever been done, to the knowledge of humanity? --- Of course, if such a guy is good enough, nobody will know - except his clients. Do we have such clients here, then let me know.
Klev, thank you so much for your pleasant experiment with me. Interesting it is, though the "French" words go way over my head. Are you refering to the Nordic mythology with "TOR", or is it an abbreviation for Tresure Oriented Research ;O) I remind you, that the really screwed, unreliable, criminal, creative, loveable assisting god in my mythology is LOKE.
Don't forget to contact me - anybody - if you hear about business-orientated information regarding this item.
Thank you all.
Best regards,
Eigil Skovgaard

[klev]

Re: Is it possible to hide a digital adjustment?

So Klev, is that a yes or a no to the original question!?

Well I can't think of any way that you could get something truly perfect for a general case. You might be able to hide the work enough to trick people on a case by case basis. There are people who have to examine things like evidence, so it would be helpful to know what they look for when trying to detect forgeries.

That wasn't as much fun to write as the first answer.

Apparently this takes further reading ... and I have still not heard the final verdict (much like the impatient customer with the Cadillac Man: stop talking about engine and transmission and tell me if the f.. car can run). Can this be done? Has it ever been done, to the knowledge of humanity? --- Of course, if such a guy is good enough, nobody will know - except his clients. Do we have such clients here, then let me know.
Klev, thank you so much for your pleasant experiment with me. Interesting it is, though the "French" words go way over my head. Are you refering to the Nordic mythology with "TOR", or is it an abbreviation for Tresure Oriented Research ;O) I remind you, that the really screwed, unreliable, criminal, creative, loveable assisting god in my mythology is LOKE.
Don't forget to contact me - anybody - if you hear about business-orientated information regarding this item.
Thank you all.
Best regards,
Eigil Skovgaard

TOR means the onion router. My other reference was to RSA encryption. I'm in an extremely silly mood today, which is reflected in my answers. TOR is actually used for some normal things, but it's also used a lot in illegal activity. It provides you anonymity under certain circumstances, and it has some downsides.

It was still fairly realistic. I mean you were asking about a "perfect" reproduction, so I answered on that basis. If I wanted to conceal something like that, I would learn about how others detect forgeries. I mean some things are super obvious.

An example would be inconsistencies in grain, which is why it would be important to try to minimize its initial appearance, presumably setting the drum scanner's speed a bit low then using post processing techniques. Correcting a mismatch in post is usually done by adding noise. It works well enough for most things. It would not hold up under close inspection.

[Eigil Skovgaard]

Re: Is it possible to hide a digital adjustment?

Klev, I know I initiated a technical discussion. Just fell off the waggon in the first curve. Hope I have not provoked the "silly" mood. I really appreciate your sharing of knowledge here. And "Tor" - I didn't know.
Thanks again.

[klev]

Re: Is it possible to hide a digital adjustment?

I was being a bit silly but still fairly accurate. If I typed the same thing in a scientific forum or anyplace containing graphics or computer vision researchers, they would say the entire thing screams citation needed. They would be right, but I was trying to give an intuitive explanation.

If I starting posting links with words like "stochastic" and "markov field" most people on here would not enjoy reading them. I have a pretty strong interest in image processing, so I learned a lot of background material to support it.

I want to add that you're looking for solutions to multiple problems. Basically you want a process that can take a developed transparency through a scanning process, then expose another transparency with the same density values. You also want to remove traces of the original film grain, so that it doesn't create new weirdness. Of course the grain pattern would be different at that point, but I think that would still be better. Issues of tools that won't leave a trace are a matter of the precision of the tool and possibly the bit depth used. You would want some tests to validate any concerns over pattern issues before proceeding to output the film again. This is all hypothetical though. I wouldn't try to get away with it.