PhotoAcute Studio Review

Many months ago I heard the term "super-resolution" and checked it out. Such software takes multiple exposures of the same scene, and runs an algorithm to enhance the details and resolution, while lowering noise in the process.


Here's an example from Wikipedia:

"An example of super-resolution processing using PhotoAcute software for Windows. The left half shows the original, blown up to twice its normal size. The right half shows the image (at native resolution) that results when the software combines nine images together and does 2x superresolution (and dynamic range enhancement)."

Super-Resolution Test

I shot 12 images of a small part of my liquor collection. (Yeah, I'm a rum snob. Didn't you notice my sig?) They were shot from a tripod at ISO 100, set at 1 second and f/5.6. I loaded all 12 images into PA, and waited a bit while it converted to DNG on the fly. I then set PA to perform its default super-resolution process.

The image load took several minutes thanks to the conversions to DNG, and once I clicked OK the initial estimated processing time was 36 minutes. 36 minutes! I'm not running a slow machine - I have a quad-core Intel Q6700 @ 2.66 GHz with 8 Gigs of RAM. Thankfully the initial estimate was bogus, and it took less than 9 minutes. But then PA took another 2 or 3 minutes to finish processing, and I had to wait before I could save the result to disk. I'm not sure what this extra time is about.


As a note, PA seems to be OK, but not great, with multi-core. During the processing all four of my cores were running, and overall they averaged at 33% utilization. It would be nice if they made better use of multiple cores. That 9 minutes could possibly have been 3 if those cores were peaked the entire time.


My 50D images are 4752x3168 pixels, which comes out to 15.1 megapixels. PA's super-resolution result was an image at 8960x6016 pixels, and 53.9 megapixels. This shows an 86% and 90% (respectively) increase in pixel dimensions, so the super-resolution process almost doubled the dimensions. The resulting image was 257% larger in megapixels.

A random image from the sequence of shot, resized to 428x285 (9% of the original size):

Noise Reduction Test

I took another 12 shots of the same subject, but this time hand-held. I had to get to ISO 12800 to get decent shutters speeds of 1/100. Again, these were shot at f/5.6, shot in high-speed continuous mode (6.3 frames per second). These images were processed with PA's default noise reduction setting.

My 50D images are 4752x3168 pixels, which comes out to 15.1 megapixels. The end result was an image at 4672x3008 pixels, and 14.1 megapixels. This loss of image size is due to the variances incurred by hand-holding. PA extracts the images where they all over-lap, and crops the excess when my handholding wavered. I lost only 80 pixels vertically, and 160 pixels horizontally.

After processing all 12 images, I then selected the first 5 images and processed them again with the same PA settings. I have to admit that shooting 12 images is a little excessive, and I wanted to compare the improvements that come with an increase in the number of images.


A single original frame, randomly picked from the 12 shots:

Super-Resolution And Noise Removal Test

PhotoAcute lets you combine many of its features in any image processing scenario - you can do super-resolution HDR, for example, or Noise Reduction at the same time you do Brightness Equalization. So I decided to try Super-Resolution and Noise Reduction together, on the same set of images used for the test above. Again, this is a sequence of 12 Images, hand-held, ISO 12800, 1/100 @ f/5.6.


Original 50D image: 4752x3168, 15.1 megapixels
Resulting Super-Resolution image: 9408x6176, 58.1 megapixels, showing a 98% and 94% (respectively) increase in dimensions, and a 285% increase in megapixels.


This image shows a 200x200 crop of one of the original images on the left, a 400x400 crop showing the same area from the PA 5-shot version, and a 400x400 crop showing the same area from the PA 12-shot version.

Moving Object Removal Test

PhotoAcute can remove moving objects from a sequence of shots by comparing them and removing things which are not in the majority of the images. I went out and visited a local parking lot (in the pouring rain), and took some shots as cars drove by. In the sequence below I was lucky enough to have some gulls fly through the scene at the same time as the truck drove through. (The gulls are the small white blobs in the shrunken sequence below.)

High Dynamic Range Test

This feature will produce a 32-bit image after processing a sequence of exposure-bracketed images. For my test, I took a sequence of 7 shots that I had exposed 1 stop apart. As far as I can tell, PA did a perfect job of merging those shots into a 32-bit image. I compared PA's results to the results from Photomatix and Picturenaut, and I really couldn't tell which of the three did a better job.

I have to admit, though, that I'm not all that familiar with such a comparison test. I've done enough HDR processing to have a very good handle on it, but my comparison test showed very little difference between the results from 3 products. Maybe it was the sequence of images, maybe it's my lack of knowledge, but they all looked fine.

Since PhotoAcute doesn't do tone-mapping there is no sense in showing the results, since the 32-bit result will not display correctly on a low-dynamic-range monitor. Anything that I could display would be the tone-mapping procedure from another program.

PhotoAcute doesn't do batch processing, so I have to wonder if this feature is very worthwhile. Many other programs (some free) do both tone-mapping and batch processing, so I'd have to say that better solutions exist if you wish to simply merge an HDR sequence.

However, don't forget that PA will let you combine features in one sequence. You could take a large number of images and produce a Super-Resolution HDR with Noise reduction and remove barrel distortion along the way. Considering this single feature alone is not really fair to PhotoAcute.


Equalize Brightness Test

I ran the same 7 photos through PhotoAcute again, this time set to Equalize Brightness rather than HDR.


Sequence of 7 exposure-bracketed images:

Annoyances

PhotoAcute Studio does have a few annoyances that I ran into. None are harmful, or even very painful, but I feel that I have to mention them with the idea of full disclosure. I have to say that everything about the interface works decently - it's fairly intuitive software- but I just found a few things to be a bit annoying.


Whenever you load a series of images you have to go back and click the check box next to each one that will be used in the processing. I don't get this at all - I'm loading the images I want to work on, so I think they should all be selected automatically.

When you maximize the PA window, it covers the Windows Taskbar.

Pressing F1 or clicking the ? button brings up the Help info in a browser window, with no indication that it did anything. It really should switch to the browser window or something. I just don't like buttons that don't appear to do anything.

It's difficult to find out what version I'm running. In Windows, this is usually under the Help...About menu, but PA doesn't have the normal Windows menu bar. Instead, the version information is buried under the Settings menu, in a tab called About.

There's no automatic update for the product or new profiles - though you can subscribe to an email "newsletter" which will announce new versions and profiles. I just stumbled upon this newsletter the other day when browsing through their web site, so I have not received an email.

PA does not support batch processing, which I think would be incredibly useful. The image processing that it performs is often very time-consuming. I'd much rather set up a number of sequences to process, and come back when they're done.

It would be nice if the program made better use of multi-core CPUs. While it does use multiple cores, my cores consistently ran at 25-35% utilization, so I would have to think that this could be improved.

The help is quite inadequate. It's enough to get you going, and I had no problems using the software since it is fairly intuitive. But some of the options (Lattice Density and Mask Precision) have no real information anywhere. There are tool-tip pop-ups for these two settings, but they give almost no information about them. Information about some of the process is also quite scarce, with a paragraph or two in the manual.


Like I said, these bits are annoying, but not damaging.


Final Words

Despite my lack of success with a few tests (like removing moving objects), I like PhotoAcute Studio and I'm glad that I own it. It has a few extremely-useful features that can improve or correct image sequences. I've been in many situations where I haven't bothered taking a shot simply because I knew it wouldn't succeed. PhotoAcute can pull successes from those situations, as long as I'm able to shoot a sequence of images.

I find its noise reduction feature to be incredible. I can take a series of shots at ISO 12800 and produce a very fine image. Granted, this requires a fairly static scene, but I know I can do it.

Its super-resolution features are pretty darn good too. It can effectively double my focal length, or double a scene's possible print size. It can pull details out of scenes that are barely legible with single exposures. I often stitch photos for panoramics or simply to increase resolution, so this feature is quite useful to me.

PA's ability to remove moving objects will come in handy some day, I'm sure. I won't expect magic out of that feature, but it will save some scenes.

Some of the other features aren't very compelling when viewed individually. But PhotoAcute's ability to combine features within a sequence make these other features useful bonus items.


All in all, PhotoAcute Studio is a tool that I'm glad that I have at my disposal. I probably won't use it often, but I'll be glad that I have it when I need it.


PhotoAcute Studio Web Site: http://www.photoacute.​com/


By the way, I have nothing to do with PhotoAcute, or anyone who works there, or any of its affiliates, I'm not getting any compensation from this review, yadda yadda yadda. I just found the software interesting, so I put together some tests to see what it could do, and that test process turned into this review. It is as unbiased as I could make it.

Great review, Scottes, of what sounds like software with a lot of potential! And, looking at the list of profiles, I even see that between my lenses and the four Canon DSLRs I have access to, I actually can put some combinations together that are not listed!

It sounds like this together with the stitching methods you mentioned put DSLR photography even more in reach of medium format quality -- imagine 5D2 files run through that puppy!

Well, thanks to this review I'm sure we will see a whole lot of POTN members jumping at this -- maybe we should expand the HDR section to include results from this puppy and the resolution enhancement!

Thanks for spending the time to run these tests and share this report!

tonylong wrote in post #8546732
It sounds like this together with the stitching methods you mentioned put DSLR photography even more in reach of medium format quality -- imagine 5D2 files run through that puppy!

Given that my 15.1 mp 50D went up to 58 megapixels, a 5D2 should hit 75 megapixels easily. Cripes. 5616 x 3744 pixels, increased 90% on each dimension, that's about 10,700 x 7100, which is 36" x 24" at 300dpi - or 60 x 40 at 180dpi.


But I have to wonder a PhotoAcute super-resolution image would compare to stitching 4 images shot with a lens with twice the focal length?

Well, given the overlap you'd really have to stitch 6 images, and I have no doubt that the PhotoAcute method is a lot easier to do than taking and stitching 6 images.


I may have to do another test. 100% crops of 24-70 @ 35mm done super-resolution, compared to 24-70mm @ 70mm single shot. Why didn't I think of that yesterday?? ??

*resurrection*

I've seen this program on the net every now and then, but didn't really pay much attention to it until a couple of days ago.

I make reproductions of paintings and other such things, and up untill now I've used to stitch shots from my 5D2 to produce 6~9k x 5~7k 16-bit tiffs, which is normally enough to make a fully usable file for a 1:1 canvas print (medium format scans were usually about 7k on the long side, so not much has changed.)

Having tried PhotoAcute Studio a couple of days ago, I'd say I'm going to have to start changing the way I shoot after having done some tests, like the example below.

HTTP response: 403 | MIME changed to 'text/html' | Byte size: ZERO | PHOTOBUCKET ERROR IMAGE

Ran some files through PhotoAcute and saved back to RAW, and then imported to Lightroom 2. It reports 84.5MP :P