It is not an easy problem to solve. The AI algorithm has to be trained and how do you do that. Feed it with the millions of Pentax images floating around on the web?
fotonor: I don't think that insufficient training material is the problem because the denoise function is working perfectly well with older Pentax raw files from K30/K50 and even prehistoric PEFs from my K100D. There must be a technical reason that DNGs from K3-iii etc. cannot be denoised easily. I tried DxO PureRaw and it shows the same behaviour. Anyway: ACR 16.1 has been released, no improvement.
The older cameras you mention, have raw files with raw data. They have a noise profile which is identical to all other cameras which output raw data. The new situation with K-3 III (and a few other Pentax models) is that they have inbuilt noise suppression. That means the raw data in the dng file is not "raw" anymore, but processed and have a completely different noise than K-50. The result of this is that the AI algorithm has to be trained to be able to recognize the noise from K-3 III and which is unique to K-3 III.
For Adobe, this means they have to train AI for a specific camera modell instead of using a generic model that works for all cameras without inbuilt noise suppression. That is not an easy task.
The "easy" thing to do, was that Pentax made this inbuilt noise suppression on raw files optional. I dont need my camera to remove noise when PC software is better at it.
I fully agree with that but that will not happen. Hopefully we still can use Topaz Denoise on non-RAW file, but that would be nice to have an official statement from Adobe because at the moment Lightroom do not display any error or warning, but just do nothing on the output.
Would be good if Lightroom threw up an error right now but there is an official statement pinned by Rikk above:
"Note: a change is required by the manufacturer before this issue can be resolved. "
So there must be some issue with the pentax files that prevents AI denoise from working on these. I have no clue what that could be but it must make the noise behave completely differently from normal. It does indeed look very differently in character than what I see from my cameras in the files posted here which all show noise limited by the physics of the image capture process which means it must be preprocessed somehow by the pentax cameras. Topaz is surprisingly bad too (although a tiny bit better than Adobe's denoise but far worse than similar files from my cameras) at removing noise from these files. This means something is different about them indeed that doesn't correspond to normal noise in such files.
@JohanElzenga Raw files from the Ricoh cameras have the same issue as the Pentax that the thread is about. AI denoise does nothing to the noise in the images. Just confirmed with some files from: https://www.dpreview.com/sample-galleries/9930153897/ricoh-gr-iii-sample-gallery/6881829042 and yeah, AI denoise does not reduce the noise in those files at all either. I think perhaps the raw files on this camera are also preprocessed by the camera before writing out so AI denoise can't do its job anymore.
you can always get DxO PhotoLab and use it's AI/ML NR and then export DNG to be used in ACR/LR ... just like with Adobe you of course will be saddled with an extra DNG
you can always get DxO PhotoLab and use it's AI/ML NR and then export DNG to be used in ACR/LR ... just like with Adobe you of course will be saddled with an extra DNG
I don't have DxO but I do have Topaz PhotoAI and it also cannot remove the noise from the Ricoh files I linked to above (https://www.dpreview.com/sample-galleries/9930153897/ricoh-gr-iii-sample-gallery/6881829042 I used the ISO 1600 image of green flowers). The problem is the files the camera produces just like with the Pentax K3 where Photo AI and other AI noise reduction software also won't work, so this issue is not unique to Lightroom.
> I don't have DxO but I do have Topaz PhotoAI and it also cannot remove the noise from the Ricoh files I linked to above
wrong bet on tools on your part... DxO can remove that noise ( DeepPrime seems to work better than DeepPrimeXD ) from those Ricoh raw files
> The problem is the files the camera produces just like with the Pentax K3 where Photo AI and other AI noise reduction software also won't work, so this issue is not unique to Lightroom.
as noted there is no problem - DxO PL AI/ML NR works on those raw files... just get the tools that actually work, stop complaining
PS: and as a bonus DxO PL comes with optics corrections for your Ricoh's... so DNG you will export to ACR/LR will have both NR and optics corrections already applied
Hi Rikk, I own a K3iii (edited) and I ahve the same issue not having the AI denoise reduction working as expected. All the sftware are updated and I can add the same issue is using .PEF file produced with the camera not only using the .DNG file. I do not know if this is relevant for the guys that are investigating but the question is : as far as you know any statement from Adobe? Any idea of solution if planned? Thanks
Bonjour, j'ai changé d'ordi et mon nouvel iMac 2023 M3 est moins efficace dans la suppression du bruit avec l'IA de camera Raw. Le précédent (intel core i5) était très lent (10 mn) mais le résultat était concluant. À présent il n'y pas de différence entre le dng original (sans denoise) et le fichier dng avec accentuation. Une explication ? La version de Photoshop pour les puces d'Apple serait-elle responsable ?
To chip in: I have the same problem with my K-3 MK iii. I would love to stay with Lightroom. I liked the denoise with my K-5 and Lightroom is easy to use but if this doesn't get fixed, I might not continue my subscription and switch to DxO PhotoLab. There I can use denoise although it is not as good as what I have seen with Lightroom and my K-5, imo.
Hello, denoise of a Pentax XF RAW file inefficient with LightRoom Classic. I changed computer and my new iMac 2023 M3 is less efficient in noise removal with camera Raw AI. The previous one (intel core i5) was very slow (10 mn) but the result was conclusive. Now there's no difference between the original dng (without denoise) and the dng file with sharpening. Any explanation? Could the version of Photoshop for Apple chips be responsible?
I've just done a noise reduction test with another camera model (Lumix dmc- FZ300) and the noise was removed in 20 seconds. So the problem isn't with the Apple M3 chip, but with Pentax's raw files, which Camera Raw doesn't interpret correctly