• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

PhotoAcute: super-resolution science?

Jeremy Jachym

pro member
I happened across this thread on outbackphoto.com: http://www.outbackphoto.com/CONTENT...nd_tonemapping/20071002_PhotoAcute/index.html

Here's a link to Wikipedia's definition of super-resolution:http://en.wikipedia.org/wiki/Super-resolution

From what I can tell from the examples it looks like it sharpens and does some local contrast enhancement, but I'd like to do some tests and see for myself how it compares. Here's a link for the download:http://www.photoacute.com/studio/download.html

I've always assumed you couldn't make more detail, but if I understand this correctly it's not a matter of making more detail, instead it's a matter of extracting more data by combining several files together.
 

Asher Kelman

OPF Owner/Editor-in-Chief
Jeremy,


This technology is also used by Reindeer Graphics for their plugins and many others. The successive frames can be used to determine what is noise and also the get an idea of out of focus elements as each image will be slightly out of alignment with the pixels so we are spanning pixels differently. I also imagine they can do deconvolution to deal with OOF issues. All in all, they do get more detail and resolution. There is so much more data avaliable for the shadow areas that there is sufficient to pull out more details.

This seems like magic, but what you are getting out still depends on what's going in and the ability of the software to accumulate values from corresponding points on a whole bunch of pictures!

Asher
 

KrisCarnmarker

New member
Hi Jeremy,

I have a license for PhotoAcute Studio (received in return for providing them with data for the Canon 100mm f2.8 lens) and have done quite a bit of experimenting.

What I can say is that is does work as advertised, but only within reasonable limits. An example of where it does not work is when you are diffraction limited. This makes sense of course, but just be aware of it.
 
Hi Kris,

same here (Olympus gear) I expect my version/instructions in a few days.

I consider this to be a useful tool in deed, what puzzles me is how that compares with the likes such as genuine fractals and qimage, theoretically it should give better results for print output.
 

Michael Fontana

pro member
Hi Jeremy,

I have a license for PhotoAcute Studio (received in return for providing them with data for the Canon 100mm f2.8 lens) and have done quite a bit of experimenting.

What I can say is that is does work as advertised, but only within reasonable limits. An example of where it does not work is when you are diffraction limited. This makes sense of course, but just be aware of it.

Kris, so where are the limits? How fast is it?

My Cam, the 1 Ds-2 ins't supported, yet...
 

Ray West

New member
Hi Georg,

I think it is something you will have to play with, see if it is of use to you. I can see it effecting the general 'artiness' of the image, giving the effect of increased depth of field, putting the detail back into the bokeh, etc. But, in other circumstances, it may be a saviour. However, I suspect you could get better results in changing to a zoomier lens, and stitching. (but then, it is easier to switch cam to multi shot mode, and just fire off a string of shots.)

Best wishes,

Ray

ps - just downloaded, and its sunny outside...
 

KrisCarnmarker

New member
Michael, I guess the limitations are all based on the fact that there is no way to increase resolution (using their method) when there is no additional info in the other images. So, OOF will (and should) stay OOF, and as previously stated, diffraction limited rez seems to be a limit as well.

As for its speed, it's not fast at all. Certainly not something you would use on all images.
 

Asher Kelman

OPF Owner/Editor-in-Chief
With multiple images, one is giving more data, plain and simple!

1. There is overlap on pixel level

2. The shadow areas have more data to rise above noise

Asher
 

Ray West

New member
Similar software to this has been mentioned before - I recollect a system for getting high resolution images from pal video frames - probably from about 15 years ago, or so.
 

Michael Fontana

pro member
Michael, I guess the limitations are all based on the fact that there is no way to increase resolution (using their method) when there is no additional info in the other images. So, OOF will (and should) stay OOF, and as previously stated, diffraction limited rez seems to be a limit as well.

Thanks Kris; , it's easy to understand, that if there is no information captured, it can' t be recovered by software; apart from inventing, but that's not what we want.

I gave it a try; far from understanding this tool well, it has some capabilities; now it's clear too, that it's not something you use every day. I compared it to Photomatix, and it's different, even one might tonemappe a HDR produced by Photoacute with Photomatix; this offering more edit possibilities. BTW: These big HDR-dngs resulting with Photoacute's HDR can't be open with Lightroom, as it has a filesize-limit. ACR works, though.

IMO, they need to take some further steps, to improve the workflow, too. And like every tool, it shouldn't be used in a overdone manner.

There will be a update for the 1 Ds-2, and for Nicolas: they have the 70 - 200 already calibrated ;-)


As for its speed, it's not fast at all. Certainly not something you would use on all images.

Yep, a improvement has to come for sure; but the few examples, I run; mostly nighshots, showed quite good results.
 

Ray West

New member
I have also had quite a play with the demo, took four or five different subjects, with 20D and 24/70 L, all hand held, and using the continuous shooting mode. For each subject I processed four or five images. ( a few hours messing. It offers no advantage, afaik, compared to printing with qimage, and possibly a noise reduction process, at sizes up to about A3, if the original image is OK.

However, it does a good job at noise reduction and doubling the resolution, imho. I am unsure what happens, if for example, you have one sharp image, and four not so sharp. It certainly alters the levels in the final, combined image, generally the combined image is lighter (it has to be so, to show the shadow detail), and the contrast is reduced.

I have no idea if they have written it in some cross platform language, it is slow enough to be written in Java, maybe they have used 'real basic'. The gui is not to a MS windows standard (oxymoron), but it may be 'Mac like', I have no idea, but after a minute or two of use, it makes no difference, but I wonder why they waste time trying to alter something that is good enough.

The demo is annoying, in that the centre of the image is obliterated with their copyright notice, (in fluorescent green and again in mauve). This prevents you from comparing the image quality of the print, and is very distracting.

I would expect it to be of use for folk who need certain sized images for stock photos, say. I managed to crash cs2, trying to print part of an image, but then, that's Adobe.

I agree with Michael's comments (above) except I haven't Photomatix or other similar software to compare it with.

In summary, I was expecting it to produce more detail than it did, but that may alter with using a tripod, and I expect the improvements are more noticeable using a non L lens.

If I get a reasonable part print, I may post some jpegs for comparison. I will not be buying it.

Best wishes,

Ray
 

Michael Fontana

pro member
Ray, no it doesn't really has the Os-X-GUI ;-)

the prerelease for the 1 Ds-2, they send me, made some ownership problems, in terms of the Unix-right-permissions.

This leads to crash the app: but when setting the permissions manually, it works. It's about the temp- items, which are fairly big.

I see a potential in that app in nightshots like the attached one: some different light sources and huge contrast of the scene make it very difficult to get the local contrast fairly done; with AcutePhoto, the image data can be calculated, due to its ehanced algorythms by reducing noise and enhancing local contrast. Here it can do a very good job, starting from 3 shots from tripod. The pict show just a scene, where Photoacute can shine.

redtower.jpg
 

Ray West

New member
sorry about the large images

Hi,

I spent some more time with it today.

I selected the base images more carefully. I only played with a set of insect images, like this below
( I took 6 shots, and only worked with the four sharpest ones - The insect was running up the wall, but paused at a mortar join, but it was waving its antennae.)

fly.jpg


The insect is about 450 pixels long, head to tail.

Using the photacute software, and playing with some of the settings, I avoided yesterday's image lightening. I loaded the result, and the four initial images into cs2, and zoomed in way beyond the actual pixel size. A screen shot is below.

head.jpg


It can be seen how pixelated the original images are at this increased size, compared with the derived image. Not shown, is the fact that the antennae tips were missing, (but I could have selected one of the original images to fix that - some settings to do with movement, which I have not tried). Also not shown, but not too noticeable on the textured stone background, is the noise reduction.

I finally printed out the insect only, on A4, and even without sharpening, it is not bad at all, compared to one of the originals at this size. A few seconds work in Irfan view, gets it quite presentable, I think with normal CS2 processing/qimage it would be excellent.

The final image resolution (I had it set to crop the edges due to misalignment - remember this was just hand held, with no particular care taken) is 6900 * 4468 pixels, cf normal 20d size of 3522*2348

My methods of testing software, is generally to push to the extremes, see what breaks, then come back a tad, to get the limits. I still doubt if I will buy this software, since I have little need, but I think for many folk, it will be a useful tool, as others have said, not to be used all the time, but it may well allow things to be achieved that could not otherwise be done. However, it will take some time to work out the settings needed for a particular image set, and then when you need to use it again, you will probably have to relearn it. Due to the file sizes, etc., and the number of calculations required, it does run very slow, and it can take an hour or too changing settings/e-rendering for just one image. (It takes about 6 mins processing the set, on my system)Then you still need your normal post processing, sharpening, whatever.

If it wasn't for the watermark, I'd most likely buy the free copy.;-)

Best wishes,

Ray
 

Ray West

New member
Hi Michael,

I think it may well be in recovering information from the shadows that it is most useful. In further experiments I have made, with a test target, it does not increase resolution beyond the lens capability, (one of the things I thought may happen, until the results made me start thinking about it). For that, get closer and stitch. The noise reduction on some subjects is good. For hand held, particularly for the critical focus sort of shot, like the previous insect, then you need to examine each image to ensure you only use the sharpest ones, and if there is movement, you have to help it along the way. On my pc it is too slow to make any of this a pleasurable experience, and the user interface is not too intelligent -e.g. if you are comparing part of a couple of images, and want to open another image, the view port changes.

I'm not sure if I have the patience to try it with a poorer quality lens. I think it has more benefit there (law of diminishing returns, etc 20% effort to get 80% result...)

Best wishes,

Ray
 

Michael Fontana

pro member
Hi Ray

>it does not increase resolution beyond the lens capability<

my only "religion": There's no alternativ to a good capture!

PA likes a lots of RAM; here it went until about 4 GB.... so faster then

>I think it may well be in recovering information from the shadows that it is most useful<
Exactly, that's my thoughts, too, especially when denoising.

They changed the ownership on the 1 Ds-2 prerelease now, I might give it Monday a try.
A lensprofile s crucial, as well

Have a nice Sunday


Michael
 

Ray West

New member
A couple of points, or more.

You can set a setting to 'native look', which gives it the usual windows gui. There are other settings, which mean something. If I adjust the exposure in the raw file, using dpp, and save it, PA ignores the new exposure setting. If I save it as jpeg, the size changes, and pa will not process mixed size files. PA does not handle tiffs. If it's set to correct for CA (20d & 24/70L) it adds bad colour fringing - it seems to overcorrect.

Best wishes,

Ray
 
Top