I am new to the forum and have specifically joined to see if any of the AI products could help me with regards to removing/improving old halftone pictures.
I must admit that I am a complete novice when it comes to any kind of photographic restoration etc, however, I have tried working on the pictures using Gaussian Blur or an FFT filter and then trying to re-sharpen them but the results are not very good.
I was wondering if AI Denoise and Sharpen might be the way forward. They are very low DPI, but will be printed out at approximately 50mm x 80mm onto plain white paper. I had wanted to post an example here but it appears that I am not allowed.
If anyone would be kind enough to offer any help or advice it would be appreciated,
Half-tone is a printing process similar to what is used in newspapers, magazines, etc.
where a series of ink dots may appear as a solid to the eye without magnification… A photograph is very different. Topaz is geared toward photographs and I’d say that the products mentioned would not be of any help in what I think you are trying to do.
Thanks for the reply Eleanor. I was just wondering if that after applying Gaussian Blur or possibly FFT that either or both Denoise/Sharpen might have brought back some sharpness to the images. I just wish i could post an example on the forum where someone with more experience and who has these programmes might be kind enough to do a little experimenting.
A Gaussian Blur might blend ink dots together as it sometimes is used on a photo restoration but its a tedious job of area selection leaving dark edges a separate blend and then blending the entire area to have all of the selections come together. The results will never be that satisfactory and much depends on the condition of the original image worked on and how important it is to you. I did a quick job of the blending to give you an idea.
This is an interesting subject for me as well. In my case, one of the things I use Gigapixel for is enlarging compressed DDS files from old computer games and flight sims. Our new computer system GPUs are capable of easily handling texture files with much better resolution now, so upsizing really enhances the look of the game.
These low resolution textures often resorted to aliasing and dithering to help with the illusion of detail that couldn’t actually be included in their small file size. What I find is that it takes some adjusting of the Gigapixel settings to strike the right balance of smoothing these into nice lines and gradients, while maintaining all of the details that they were trying to represent. It’s always a compromise, but I can usually find a good balance that works when you’re back in the game.
Lately I’ve been upscaling to 4x in Gigapixel with 80% suppress noise and 40% remove blur, smoothing anything that still needs help like residual jagged diagonal lines in PS, and then downsizing again to 2x to “homogenize” the final result a little more. All of this takes converting the DDS file format to TIF for enlargement, and then back again to DDS for the game, but that only takes a minute.
Thanks once again Eleanor its appreciated. I just wish i could upload one of my examples,
Thanks for the reply Andy. Thats actually not a bad result. I did wonder if trying this route might work but being a complete novice i wasn’t sure.
I just wish i could upload one of the pictures so that someone might experiment and i could see if Denoise or Sharpen etc might improve things,
Just out of interest does anyone know how long it is before new members can add files to posts?
I believe it was 5 posts Peter but the original guidelines seemed to have disappeared when the site was redone.
I’m a raving fan of Topaz filters, but halftones as others have said is different than photographs. To get the halftone out, I recommend the Sattva Descreen filter. I have the pro version. It may not be worth it to you if you only do the occasional scan. I collect vintage images and they are occasionally in halftone, so it was worth it to me. Then, you could use Topaz to clean and enhance if needed.
Thanks Eleanore, if that’s the case then i should be able now.
Hi Inicole, thanks for the recommendation the Sattva Descreen filter looks interesting, quite expensive though, at 75 Euros for the Pro Version, it would be interesting to see how it handles some of my pictures though.
Just tried to post a picture again but it still won’t allow me to do so, maybe the minimum is ten posts?
This is a typical example of the halftone picture that i need to address,
Got some information for you…Discourse, the software used on the site has to due with trust levels…https://blog.discourse.org/2018/06/understanding-discourse-trust-levels/
You were still considered a “Basic Member”. Thanks to AiDon, one of our moderators, you have been upgraded and should now be able to upload.
Thank you Eleanore, that is much appreciated, and thanks AiDon.
LOL. See you made it. I think one of the first things you should do before trying is increase the exposure…lighten it up.
Yes your right it is a bit dark.
DeNoise AI did nothing to remove the halftone pattern.
I tried the approach in the video below, and it revealed that there is some large scale blotchiness from damage to the paper original. Unfortunately as you enlarge it with Gigiapixel the blotches becomes quite noticeable. Gigapixel also created some color variations that needed a desaturate to clear up.
Sharpen AI can be used once you’ve eliminated the screen patterns by blurring them, but it brings out the blotchiness even more.
Thanks for taking the time Andy, it is really appreciated. Looks like this is going to be a difficult one.
Just so you are aware, this is actually a photograph of a newspaper halftone picture. I was allowed to look and copy some parts a newspaper archive, many years ago now and there is no chance of scanning the originals etc. This is what I have to work with and so I will have to make the best of it.
Inicole has kindly offered to run the picture through the Sattva Descreen filter for me, so depending on the results of that maybe one of the Topaz filters could then be applied more effectively.