• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Pretty neat noise removal technology for images

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,054
That looks impressive. The one other I've seen people imply they could do, but nothing to my knowledge has come of it, is fixing out of focus photos. If you could focus and denoise bad photos that would be fantastic.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,280
Location
Gold Coast, Queensland, Australia
They say "train a de-noiser without clean targets". How then does the algorithm know what a koala looks like? There's simply not enough information in that picture to get anything other than a slightly textured outline.

It must have been trained on thousands of shots of other Koalas and the particular way they hold a branch, how many toes they have and what are the type of leaves in the eucalypts which koalas sit in.

Very cool demo though. :)
 

Wombat

Master Contributor
Joined
Nov 5, 2017
Messages
6,722
Likes
6,459
Location
Australia

L5730

Addicted to Fun and Learning
Joined
Oct 6, 2018
Messages
667
Likes
434
Location
East of England
Oh great, so we can't even watermark our photos for use online anymore?
I suppose that has been an issue for a while anyway, skilled photoshopers knock out the watermarks all the time. Never upload in printable resolution is the trick.
I've used a number of different noise reduction software and one thing that is common is that with synthetically generated noise, they work really well. It's a repeating pattern and can target those bits easily. An image captured from a sensor is not a repeating pattern of noise, it's all ever so slightly different. In the past I have found that NR is needed on a case by case basis, and Neat Image, Noise Ninja, Topaz and Adobe's own Camera RAW all seem to have their own quirks that do a better job than another on a particular image - this takes too much time to use them all and compare case by case.
Is the image for screen or print? If it's for print, we can get away with a lot more noise than on screen, even to the point where some prints look better with a little added noise, it seems to make them sharper on paper.
 
Top Bottom