CCC Forum

We all have them. Images which are otherwise OK except they are not quite sharp. Autofocus did not somehow work properly, we missed where focus should have been, or image stabilisation did not kick in as expected. Mechanical systems are not always reliable, especially in the hands of people. For a while, software developers have been trying to do the impossible. Produce a sharp image from a blurred photograph. The impossible? More like one of the holy grails of photography. We might not be quite at that stage yet, but if the article at the link below is to be believed, humanity is getting closer. The author, Ctein, knows what he is doing better than most. Topaz Sharpen AI might be the best way yet o rescue otherwise imperfect shots.

https://theonlinephotographer.typepad.c ... sense.html

Best of all, the software is available on a free trial and currently on offer at $59.99. If it does not do what it claims, there is no need to buy it.

https://topazlabs.com/sharpen-ai/

As Ctein says, perhaps the only problem the application cannot resolved is how to retrieve those lost masterpieces which we so carelessly discarded. We believed the hype that it is impossible to get a sharp result from a soft capture.

Back in the day where when our images were captured on a piece of celluloid, there would be no way of changing sharpness; however with the advent of digital our images are just a bunch of one's and zeros spinning on a disk. So it is only a matter of time before they produce the software that is able to analyse our image compare it to a database of similar images, then go on to rearrange and add to your ones and zeros to give you your sharply rendered image. Personally I think it will be awhile before we see this working as we would want it. Look at the clone tool....

Franke07 wrote:Look at the clone tool....

The Clone Tool has some uses, but simply copying pixels can result in a crude effect. I quite often use the ACR Heal tool to fix minor blemishes, especially with a bit of feathering applied to smooth the transition between actual and recomputed areas of the image. Sometimes it works well, sometimes not.

For quite a while, the full version of Photoshop CC has been incorporating functions which attempt to modify images based on context. There has long been a tool to resolve out of focus images which is OK in an emergency, nothing more though. Sharpen AI looks a though it could be a better option. More impressive are features such as Content Aware Fill which attempts to recreate elements outside of the captured frame. I find it useful on those occasions when I want a bit more space around the main subject than I allowed at the time. Admittedly the results are mixed, but when it works, wow! Better photographers than me get it right every time, of course.

Incidentally, it is worth scrolling through the comments on the TOP article to which I linked as Ctein elaborates on the workflow he uses for Sharpen AI. He finds that it works best on unmanipulated images. He also suggests using layer masks to include those elements of the image I have interpreted that as:

• Create 16 bit TIF from the Raw capture without making adjustments
• Pass the TIF to Sharpen AI
• Continue processing the image in your photo editor of choice
• If ACR is your normal method of processing images, continue in Photoshop with the Canera Raw filter. It will probably not be necessary to add any more sharpening to the image.

I had a look at the Topaz link and there are some great examples on it. One shows a ceiling fan with obvious movement and after processing is static, thats mightily impressive.

On a related note an AI company has produced 100,000 AI generated portraits for royalty free use. Plus as the "people" dont exist you dont need a model release form! Some of them look very realistic - https://petapixel.com/2019/09/20/this-c ... -for-free/

Further to Frank's comment, Adobe continues to add more "AI" style technology into Photoshop. An updated Content Aware Fill function, which adds pixels not present in the original capture, will be available shortly. Story from DPReview, link below.

https://www.dpreview.com/news/125933504 ... ure-update

Well, I bought it. Put my money where my keyboard is.

I found a discount code online which gave me a further 15% reduction, so I ended up paying $51. I have configured it as a plug-in for Lightroom, which is where I reckon it will be most useful, but I will probably also set it up a in Photoshop as well and use it on a sharpening layer when required.

Having played with AI Sharpen a bit more, I can confirm Ctein's finding that the Stabilize mode has yielded the best results so far. Looking through the comments to the original article on TOP, someone has posted an example of what Focus can do, but that option has not worked for me. So far, at least. Now, here is the strange thing. AI Sharpen does something to images which are already sharp and brings something more out of them. I suspect that it boosts the micro contrast. I have already provided an example in the post for the outing on 19 October, without discussing my processing. I have repeated it below, although it a low resolution JPEG does not really do the shot justice.



The other thing I have experimented with are the GPU settings. Like a lot of applications, Lightroom included, AI Sharpen can hand off some of the processing to the processor in the graphics adapter, otherwise known as a GPU. Unlike the main CPU, which is designed for a range of copying requirements, the GPU is dedicated to image processing and can be quicker in some circumstances. My main computer has the previous generation of Intel's Core I-7 CPU and a midrange graphics card. I am not a gamer, so the latter is not where I needed to spend the money when specifying the system. AI Sharpen allows the amount of processing handed off to the GPU to be configured. When set to "medium", the main CPU was utilised at 100% and the GPU ran at around 40%, Set to "high", the GPU was maxed out, with the main processor running at around 30%. In this mode, the rendering time was longer.

Another vote for Topaz Sharpen AI. As the report says, it is not suitable for every image and there needs to be a reasonable amount of detail available for the application to work on. I have been reviewing my recent images, including those from The Great River Race on the Thames in September which I have not had the time to look at before. Of course, the best one was a bit out of focus but Sharpen AI easily recovered it. It is quite miraculous when the circumstances are right. With Black Friday coming up, maybe Topaz will have an offer?

https://fstoppers.com/post-production/r ... ble-430466

Topaz is discounting its products over the Black Friday weekend. Sharpen AI is $79.99 and there is a bundle which comprises all of its offerings for $299.99. Unlike other software vendors, Topaz Labs seems to provide updates without further charge for the life of the application, so it is a one-off purchase. Better still, any of the widely available 15% discount codes can be applied to these prices. Take a look at the review from Northlight Images, which is useful source of information for many things photographic.

http://www.northlight-images.co.uk/revi ... harpen-ai/

https://topazlabs.com

How about another point of view. Do we really only need sharp images ? I have been using my Camera Raw system for producing slightly diffused images, especially for portraits of older people, which I am about to present on Wednesday as a print. Also landscapes you can only use it unfortunately with a very strong image colouring although I am experimenting still. you probably ave seen one of my experiments of which I got mark of 7. trouble with judges they only like what they are used to, not many think laterally.
Anyway I am sure sharpness is more important, but to me its also worth thinking about diffusion.
Good luck with your projects.
Wally