Review: Topaz Sharpen AI
Topaz Sharpen AI review
Software offers shake reduction and sharpening
Topaz have just announced a new software package for sharpening photos.
The software offers both sharpening and shake reduction. The software runs at its best if you’ve a fast modern graphics card but will run under ‘CPU power’ if need be.
Keith has been testing an early version of the software to see if it could augment some of his long standing software choices.
People who purchased Topaz Labs Infocus should get an Email with a free upgrade to Topaz Labs Sharpen AI.
Sharpen makes use of similar machine learning techniques that Topaz have applied to some of their other software. This involves training software with many many photos with various degrees of problems and using the results to drive correction software.
Sharpen AI addresses sharpening in three ways. One is simple image softness, another is out of focus softness, and the third, which interests me the most is shake removal.
The software I’m testing is a pre-release version, so expect minor changes.
Essentially, you open a file, choose the settings and it goes off and processes your image. The software really does benefit from a fast graphics card.
The default view shows a small part of your image zoomed to 200%.
Looking for slight camera shake, noise and focus errors is about the only time I really look at images at this magnification.
A (moveable) split view helps compare what you started with, with the results.
[click to view at full size]
There’s an RGB histogram and HSL vector display available for the currently loaded image.
The vector tool is not one I’m familiar with using.
Yes, my photos sometimes have a bit of camera shake…
Partly it’s because I like shooting hand held, and partly because higher MP sensors are more sensitive to showing very small amounts of shake, I’ve long accepted that a few photos I take won’t be as sharp as I’d like. I happily use a tripod for much of my architectural and industrial work, but sometimes you just can’t use one. Good technique (and multiple shots if possible) will always help, but a few pixels of shake has never been the end of the world.
Here’s an 11MP jpeg from the camera (Canon 1Ds in this case).
[click for full size]
Running the software gives me a version minus the shake.
A view at 200% magnification gives a good feel for the processing.
The default settings will make a good first attempt at reducing shake, but you might want to tweak the blur amount.
See these three versions at slightly different settings. You may need to enlarge the images to really see changes
Differences tend to be slight, but before spending too long tweaking sliders…
Remember that this is at 200% magnification
The noise slider may help with noisier images where the shake correction can emphasise noise in some areas.
The grain slider will add noise ‘grain’ to your image – this will sometimes reduce a slightly artificial ‘digital’ feel to fine detail. It’s not a “What would this look like shot with Tri-X’ type of noise – there are many other bits of software that do a good job there.
Where the shake reduction does a good job is with non-linear shake.
In this example look how the curved path of movement on the left side has been corrected on the right.
For some images you may not be able to decide whether they just need a bit of sharpening or whether they have a bit of shake in them.
These three examples suggest that there is a tiny bit of shake (200% views from a 21MP 16 bit TIFF image).
Once you’re happy with how the image is going to be processed, it’s worth checking in other areas, since blur/shake is not always even.
The file can then be saved in a format of your choice.
You can specify colour space as well.
Sharpening and Focus
The sharpen/focus tools apply a different style of image sharpening. Once again there is a lot of calculation going on.
It works well on some images and less so on others. I’d likely use Sharpen AI as an an adjustment plugin in Photoshop, where I could duplicate a layer, process it and brush in parts of what I wanted sharpening.
I used this technique with AI Gigapixel in a recent article about making large prints from low megapixel images.
Once again take care with the sharpening display, since a view at 200% can give a very biased view of sharpness.
Usefulness and limitations
Do note that I’m testing a pre-release version of the software – download the free trial if you’d like to try it with your own images. Note that on a Mac it needs at least MacOS 10.12
It’s always* going to help by starting off with a sharp image, so there will be limits on just how good the sharpened and shake corrected version will look. However, sometimes the image you want is the one that’s got a few pixels of shake.
*I use the term ‘always’ more carefully these days as I see advances in computational image processing.
The shake detection coped well with up to 10-15 pixels shake, especially more complex movements that one of my favourite tools (Focus Magic – since 2004!) would not address. Beyond a certain amount though, the blurring was just not seen as shake.
One of the things I found when first testing Topaz AI Gigapixel was that preparation of the image you put into the software could make a visible difference to the output. If the images you want to test with Sharpen AI are from RAW files then try processing them first to TIFF files but with all sharpening turned off. Depending on the camera and RAW processing software, it’s possible for noise reduction and sharpening artefacts to appear.
Of course, all this work has to depend on just how important the image is to you…
The software is available as a free 30 day trial
Never miss a new article or review - Sign up for our occasional (ad-free) Newsletter
Enjoyed this article?
All the latest articles/reviews and photo news items appear on Keith's Photo blog
We've a whole section of the site devoted to Digital Black and White photography and printing. It covers all of Keith's specialist articles and reviews.
For All about using tilt and shift - articles/reviews about tilt/shift lenses
Articles below by Keith (Google's picks for matching this page)