Bart,
Sharpen AI is promising, but I have been unable to find any documentation on how to adjust the parameters or in what order to apply the corrections if one is using more than one of the three algorithms. Any guidance would be much appreciated.
Hi Bill,
There is a very minimalistic "Product Tour" when opening the Stand-alone version (it can also be called from the Help Menu). Opening the application, with Photoshop closed, should add the app as a plugin to Photoshop.
There is not that much to tell.
One uses one of three modes wich chooses between three different AI models.
- Sharpen sharpens the most sharp pixels in the image.
- Stabilize does the same. but at the same time attempts to remove motion (subject motion or camera shake) blur up to 10 pixels long in any direction.
- Focus attempts to refocus images that missed critical focus when the shot was taken (defocus creates a different type of blur/PSF).
So in principle, it only takes one of the three. Of course one can try to save a shot and reload it for another pass, but that's not how it's intended to be used.
There is currently a risk that
Focus will pick up details that do not belong to the plane around best focus, like parts of the background. That's why it has been requested to add some sort of masking control. When a lens design e.g. creates spurious background resolution, it will be mistaken for OOF blur. The folks at Topaz are looking into that. For the time being, just use a masked layer to hide any overzealous attempts at refocusing.
Because it takes a relatively long time to process the entire image, the preview area only is processed to save time, when you hit the "Update Preview" button. So zooming in will require fewer pixels to be processed and that speeds things a bit up. When the preview is moved, or the zoom percentage changed, one has to hit update again. The space bar, or the icons in the top menu bar toggles between processed and original image view. There is also a Split-view preview possible. Only when Saving the image will the entire image be processed, which can take some time.
With the controls, one tries to match the amount of blur, or noise, in the source image. More is not better, but a match gives the best results. That takes a bit of guessing, but one gets the hang of that after a bit of trying on one's own images. The
Add Grain control can add a bit of uniform grain back to the image to avoid too smooth gradients. To learn, one can start at 0 and update (which only does a little) then 10 and update, and so on to try and find a sweet spot which is just enough to not produce nasties, like halos or stair-stepping. I usually see something useful happening at 20 or 30, or at 70, but that depends on image content. Undo/Redo steps through the changes.
Hope that helps.
Cheers,
Bart