I just learned about de-convolution technology for recovering detail from apparently blurred photos, and saw this service that does it commercially. I thought it might be interesting to the BF community, in that their software seems to be able in some situations to recover a surprising amount of information from what look like very poor photographs.

http://www.picturesolve.com/

If it hasn't been tried before, it could be interesting to see what they could extract from a blobsquatch photo, or even some still frames of Patty.

Anyway, just thought i'd post the link, and apologies in advance if this has been discussed already.


-robin

No kidding, that's awesome.

Is this a product?

It seems like an area of ongoing research, but somebody is evidently trying to commercialize it as a service. I'm not sure what software they are using, but it's possible that it's custom developed.

Here's some more technical (and i do mean technical) info: http://people.csail.mit.edu/fergus/research/deblur.html

Hey thanks. Good information.

I have zippo technical skills here but it is a very interesting idea.
Is anybody else on this board familarized with this kind of program?

Huh. Interesting. I'm a bit skeptical on some of the examples, but very cool if indeed it works as shown.

Very. The problem with this kind of image processing is that the images must be very clean and not highly compressed for the really impressive results to occur. But if you could do that, then not shaking the camera in the first place is the best idea.

Another issue is what digital cameras do to take a picture. Some take 3 passes on the picture for RG and B, then dither that into a single buffer, then perform JPEG compression after that. That is why it takes so long for the picture to snap. This is changing, with the new 3 CCD models, and other quantum chips that work more like the human eye.

You can also set your camera to use higher quality JPEG compression but it costs more memory usage. These type of digital and algorithmic improvements sometimes make it to the consumer level, and then things can really improve. So the single picture stabilization algorithm will probably end up as a photoshop filter...

I submitted a request for the use of the code on behalf of SRI. I filled out the request using SRI as the research organization.
I don't expect to receive an answer..but you never know.

This isn't anything new. Deconvolution filters have been around for +30 years and Photoshop already has many 3rd party plug-ins for this. You can find various source code for deconvoluting algorithms on the net, and in numerical publications. These algorithms are very powerful for limited applications.
Here's a tutorial explaining the process.
Here's a free deconvolution plug-in for Photoshop

Howdy folks...

Image Analyzer has the Deconvolution filters built in...And it's free :new_thumbsupsmileyanim: ...

http://meesoft.logicnet.dk/

Does anyone know of a publicly-accessible pool for unedited, but post-digitized stock footage of either the Patterson-Gimlin film or the Freeman footage, without having to raid the LMS supplementary sections (namely because the overprint can impact the Fourier space processing, and because it was really unnecessary to slap the obvious on those DVD inclusions)? I might want to try a hand at a deconvolving pass on either of these as a painstaking summertime project. Actually, two or three test frames is fine for now if anyone has something like that. It might be worth posting before, after (prior cleanup attempts) and after (post-deconvolution) of a handful of test frames to see what others here think. Cheers.

Did you ever get a response?

Actually, I may as well go with the DVD transcription on LMS. My concern with that is that it's yet-another generation signal degradation: rather than generation loss from the film transfer of a film transfer of an original print (I think the DVD ripped from a third-gen source for the transcription, yes?), the last generation loss is from the lossy compression algorithm of MPEG-2. Given the way deconvolution works, digital artefacts from this compression can significantly alter the interpolative calculations, resulting in some apparently weird visual effects.

Anyhow, so has anyone else here had an opportunity to run test frames/photos/etc. using deconvolution steps? If so, were they in the domain of the available sasquatch footage publicly accessible? And if so, did anything productive come of it?

That's all. Thanks!

There are different copies of the original film, and as far as I know DDA used his photographic expertise to capture the best copies of the images that were later stabilized by MK Davis. The copyright and ownership of this work is murky at best. As far as I know these images are not publicly available.

There is much, much more that could be done with this film, but consumer software packages aren't going to get the job done. There is a big difference between someone who likes to use Photoshop to fool around with the images and the things that could be done with a purpose built graphics pipeline for edge enhancement, brightness and contrast adjustment, deconvolution, and stability.

There absolutely could. Deconvolution and other image recompositing algorithms on a generational descendant copy are no better than a stop-gap at this point. Given what is accessible generally, however, I wanted to eliminate that route while the estate of Roger Patterson(yes?) ultimately decides what to do about the restoration (followed by digital processing) of the original print. It's possible that that kind of decision might not happen for years, if ever.

That said, its utility might become, at most, historical and even nostalgic should other advances in this area of research yields a physical specimen or, barring that, a piece of video footage (hoax-eliminated, that is) that reveals tremendously more than what we've seen up to now. Its historical value is priceless, but at this point in time, it's a detriment to the research arm of this field to have the matter in legal flux. Meh.

That went in a huge tangent, sorry. I guess my point was this: find the best generation readily accessible and work with those contents to produce something that might (only might) be the best material usable for the time being.



OK. Here's where my own conjecture is offered -- though before I go further, I largely agree with you on that. :)

I have found over the years that an individual's means of owning and tinkering in (often the newest version of) Photoshop is less significant than their ability to work with the complex options for that tool. Personally speaking, I have been working with Photoshop since v2.5.1 in 1994, and in high-production marketing environments (which only qualifies that a lot of images have passed before my eyes over the years).

All these years and applied uses later, I know that I will never "master" it (or even the vector possibilities of its CS sidekick, Illustrator). In fact, it's the kind of robust tool that, left in the hands of the lay user, exaggerates what they think they can do with it, while for the few handfuls of people approaching what I'd reckon as "mastery" (but more like "magic" in terms of the work being so impressive that you ignore how they did it and enjoy more that they did it), even they stumble into new combinations of uses for Photoshop -- and not necessary the latest, "greatest" version of it, either.

In short, the hardware/software matching is valuable to have, but few have the means to work with ILM-like toys. Given the best of what we have, maybe a better stop-gap restoration might be of utility -- or even for the muddy Freeman footage (which as you're aware is muddy for a totally different reason). Maybe it wouldn't be better, but I wouldn't mind trying. Assuming I find the time, I'll see what I can do with the LMS footage.

Thanks for reading. I hadn't intended this to be a long response.

While the photographic capture of the original images is excellent, the currently available digital presentations of that work look like amateur hour. These images deserve professional treatment for historical purposes, and should be treated as such.

I would like to see a "PGF Explorer" application that allows this footage to be interactively explored and visualized, and that would also protect the intellectual property involved. There should be a reproducable chain of graphical transformations that can be tweaked by the end user. On the input side, they should be able to specify brightness, contrast, edge enhancement, deconvolution filters, and stabilization parameters. On the output side, we need color, depth, scale, zoom, and speed. A program like this could also serve the purpose of preservation in digital format for future generations.

Quite. Any one in particular worthy of the "best amateur attempt to look like an amateur" trophy?



Point, echoed. Just the very nature of celluloid stock is that it brittles, fades, and discolours over time, particularly so if handling is less than optimal for long-term storage. This alone would at least be validation for a digital transcription from source -- even if the digital content is left under lock and key. I'd even be willing to put a few dollars towards forwarding a kitty to the estate just to have this media digitally captured, frame-by frame, just for the sake of archival and preservation. One day, it might be something appreciated by those who use that as a reference point.. It doesn't even have to be retouched -- just as long as there's a raw digital video archive, while finding a storage means to preserve the original stock.



So in other words, a self-contained application which enables the end user to modify, tweak, shape, bend, filter and examine to their heart's content, saving perhaps the parameters they just used to effect their results, and sharing the file with those steps built in. Kind of the way Actions work in Photoshop. It's a grand idea, though the engine itself would be what's critical here; the media within that explorer engine can be swapped with any media file, really.

On another thread (and changing thoughts for a moment), someone posted photos of their journey to Bluff Creek last October. One photo carries significant enough reference content to serve as a colour balance target when restoring the hue balance in the Patterson-Gimlin film. That palette can be quite useful here should a interim colour restoration occur as part of a bigger process.