PDA

View Full Version : Has anyone tried "PhotoAcute" Studio with a "consumer" scanner, such as the V-750 Pro



yaredna
10-Dec-2007, 20:24
I have looked at the description of the algorithm behind "PhotoAcute" Studio, but I have not yet tried it. I was wondering if any of the digital experts on this forum has tried it with a scanner. The slight "mis-positioning" of the film on the scanner in successive scan will allegedly allow the software to derive computationally additional resolution.

You might want to check their web-site for additional information:
http://www.photoacute.com/studio/index.html

The way I understand it, in simple terms:
. We need to scan the same film 5 to 10 times. May be try replacing the film on the scanner plate, allowing for slight misalignment between the different scans
. Feed all the scanned images to the "PhotoAcute" software
. The software will then "crunch" a better resolution image.

I have not purchased the software yet, but they seem to have a downloadable trial version. If no one has tried it before, I can volunteer and give it a try with my Epson V-750 Pro, and publish the results.

Thanks
Yaredna

Witold Grabiec
10-Dec-2007, 20:42
This is like straight from astrophotography and the widely used photo-stacking. BTW, I don't think they imply deliberate camera movement (as you seem to suggest), in fact they say "try to shoot steady, although tripod is not required and no special precautions for shake-free shots are needed". They're marketing it for cameras only, scanner is not the same thing, but it's probably worth trying.

Kirk Gittings
10-Dec-2007, 20:46
It is not unlike some other sofware out there, perhaps taken a step further. It is worth looking at. I emailed them a question about its possible value for scanning.

Witold Grabiec
10-Dec-2007, 21:02
Here is one review (http://www.outbackphoto.com/CONTENT_2007_01/section_hdr_and_tonemapping/20071002_PhotoAcute/index.html) of PhotoAcute (scroll down a bit) and it appears that camera/lens profiles are required for it to work properly.

And here is a word on the technology (http://en.wikipedia.org/wiki/Super-resolution) from Wikipedia.

And how about some "smart" paper on the technology (http://www.soe.ucsc.edu/~milanfar/SR-challengesIJIST.pdf) ?

And more here (http://people.csail.mit.edu/billf/papers/TR2001-30.pdf), this one looks really smart. Anyways, I'm gonna stop here. As I suspected this has been utilized in astro-imaging. Technology isevolving and holding a lot of promise. Google "super-resolution" to see what's there.

yaredna
10-Dec-2007, 21:32
OK.

So I took the time, and I did a simple test. One of the posts above mentioned that the software requires a couple camera+lens listed. That was my surprise. I did however work around it, by selecting a Canon EOS DSLR 1Ds Mk II, and a 50mm f/2.5 macro lens. Just to see if the software is still capable of doing something.

I scanned a Velvia 50 (the new version of Velvia 50, shot taken 2 weeks ago in my backyard), using my Epson V-750 Pro. Resolution: 2400 dpi, no sharpening.

I then selected a small portion of the photo, to make the test faster. I enclosed here below a reduced version of the entire photo, all darkened except for the area that was tested.

I then scaned it 8 times, and loaded the JPEG into PhotoAcute. Tricked PhotoAcute to believe this is a Canon DSLR. Processed for 2x superresolution.

The results are attached. I included a screen capture from the software, the photo as saved by the software in 2x resolution, and one of the 8 scans.

I am not an expert in scanning, so I will not comment on the results. Anyone wants to comment, please feel free (I meant: not comment on my focusing capabilities with the view camera :) ).

Oops, I needed to crop the files, and save in lower quality JPEG to fit the attachment criteria.

Thank you

yaredna
10-Dec-2007, 21:34
I also resized the output file down to 1x (50%) using bicubic sharper (debatable choice). Resulting file attached.

Question: is this simply a glorified sharpening algorithm, or does it really resolve more details or texture ?

Regards
Yaredna

Witold Grabiec
10-Dec-2007, 21:42
yaredna,
Could you post "straight" results (no manipulation) outside forum with a link to it? If you have no web space, email me those directly and I'll do it for you.

Having read a little bit on the technology, it is little else but a sophisticated algorithm that searches patterns is subsequent photographs and figures out where the extra detail is meant to be. From your quick scans it looks (to my eye) no better than the Fractals resizing.

--------------------------------------------------------
Just to clarify my Fractals reference: I meant the quality of detail, but quite impressive nevertheless, especially considering the "quickness" of this test.

Kirk Gittings
10-Dec-2007, 21:58
I assume we are looking to compare the third one in the group of 4 and the single below with the watermark? The improvement looks like sharpening to me as there are some feint halos around some of the finer branches, but there may be something here worth some more investigation.

Witold Grabiec
10-Dec-2007, 22:38
OK, here are images as YAREDNA got them.

Acute001.jpg (http://wjgrabiec.com/LF_forum/Acute001.jpg)

Acute001_08.jpg (http://wjgrabiec.com/LF_forum/Acute001_08.jpg)

Photo_Acute_Screen_Capture.jpg (http://wjgrabiec.com/LF_forum/Photo_Acute_Screen_Capture.jpg)
-------------------------------------------------------------

Two more added:

Acute015.jpg (http://wjgrabiec.com/LF_forum/Acute015.jpg)

Acute015_21.jpg (http://wjgrabiec.com/LF_forum/Acute015_21.jpg)

And I agree, not much over sharpening. However, let's see if a better test can be provided. This software comes up with camera profiles, and that could be part of the problem. However, super-resolution is still "just" a mathematical problem and how it's modeled. According to the "smart" people it has future potential. By then we'll be driving cars on tap water though.

Kirk Gittings
10-Dec-2007, 22:49
Seems like a pretty modest gain, which looks very much like what simple sharpening would do to me.

yaredna
11-Dec-2007, 17:50
I emailed the support team at PhotoAcute and inquired if their software could be calibrated for a flatbed scanner. I shared with them the link of this thread. If and when I receive an answer, I will share it with this forum.

Thank you

Kirk Gittings
11-Dec-2007, 18:35
If a profile really shows more promise than we see above, it would be useful to be able to order custom profiles for different lens, film, scanner combinations?

yaredna
11-Dec-2007, 19:47
Kirk,

I believe different profiles would be required for different scanners, and for each scanner, different scanning mode (dpi, reflective or film).

I do not believe that the fim type (Velvia or Provia, for example), Focal lens, aperture or other set-up do require different software set-up (different profile).

But then, I might be totally wrong.

I totally agree that if the results allow a gain of, let's say, 50% resolution (1.5x), making the Epson 750 a "3200 dpi" scanner, it will definitely be worth the investment.

Regards

Kirk Gittings
11-Dec-2007, 21:28
Different scanner profiles are available in SF and high end scanner software for different color and b&w negative films now. To completely profile a scanner you would have to take into account grain size and characteristics of the film etc. a factor that may be akin to incorporating sensor array data and pixel density for profiling DSLRs. Right now it doesn't matter allot because the prosumer flatbeds cannot resolve grain, but if you were able to come close to resolving grain, the characteristics of that grain would start to become important as it is in drum scans to avoid grain enhancement etc. If you are going to bother profiling a major piece of your workflow, why do it halfway?

yaredna
13-Dec-2007, 10:30
Here is the answer that I received from AcutePhoto's team. Do any of the experts on this board know of a way to generate the most accurate film target that they would need? This would amount to following a similar process to that used, for example, by Kodak to generate their IT-8 targets, right?

----


Thank you for your interest.

In principle, it is possible to create a profile for a scanner.
The profiles are created using the sets of pictures taken with
particular optical system and the reference picture. For example, for
generic cameras it is possible to display the reference picture on a
monitor and take shots of it by the camera being profiled.

For the scanners the question is - whether it is possible to create a
'reference slide'. If the resolution of a slide is determined by a
film - it is possible to take a photo of the sample picture and use it
as the reference slide. But if the resolution is determined by the
camera optics - the profile will be created not for the scanner but
actually for the combination of scanner + camera + lens.
Of course, it is possible do the following - take a photo of the
sample picture, use it as the reference slide, create the profile and
then try applying the same profile for the slides taken with other
lenses.
Another possible way: scan the reference slide with a scanner that
has much better resolution than the target scanner (four or more times
as much) and then use the high-resolution scan as the reference
picture. We mean effective resolution, of course, so if your scanner
gives 1800-2400 dpi it would be enough to use the one giving 4800 dpi.

We are ready to experiment with scanner profile creation if you can
provide us with necessary sample scans. As this is to be an experiment
- we will do it at no cost.

Sincerely,
PhotoAcute Support

-----

Asher Kelman
14-Dec-2007, 13:19
Thanks for the effort. Posting to get email updates! :)