Everyone is on this bandwagon right now, so why not me? Just a few days ago (April 2023 for you people from the future), Lightroom added a feature most of us have been dying to have for ages – better noise reduction! There are other things that are cool like curves adjustments for masks, but this is huge and can ultimately save you money if you don’t have to pay for 3rd party apps to do this. Just when I almost pulled the trigger on Topaz Photo AI, Adobe rode in a white horse, but have they saved this damsel in distress? Let’s take a look.
After watching some You Tube videos about this and deciding that the new update was stable for Mac Ventura, I started experimenting. Mainly I picked high ISO photos that needed substantial cropping – the worst case scenario for a Micro 4/3rds shooter. The new denoise function is accessed in the same places as the old and will produce a new, separate DNG file that can be stacked with your original RAW file (no jpgs, TIFFs or PSDs). There is only one slider to change the amount of denoise it will do. I find around 60-70% works well for the photos I’ve tried so far. It doesn’t appear to make a difference if you apply sharpening before or after. I didn’t apply clarity or texture before denoising, but applying them after works just fine. I won’t post everything I tried, but here is an example of just how good this is right out of the gate –
That is with Lr denoise with the default level of sharpening (+40) with no texture or clarity, but with Photoshop’s Smart Sharpen applied to only mom’s face. I like Smart Sharpen because it can be finessed a little better than in Lr. Plus I can use masking to apply it to only the areas that need it. I sometimes work with the Detail panel in Lr masking, but it isn’t as good as what I can do in Photoshop. I suspect the way it sharpens is done differently, but I don’t know so let’s put that aside for now.
Here is a look at the same file with Lr sharpening applied globally, but restricted with the masking slider in the Detail panel –
There’s a bit of a color difference as well since I didn’t apply color correction or saturation exactly the same in Photoshop where I did some distraction removal.
And finally here is the original RAW file –
Not the worst noise ever because it wasn’t badly underexposed, but it wasn’t really useful either and I’ve never bothered to process it.
Here is an example of the best I could do with what Adobe now calls Manual Noise Reduction & Sharpening applied globally, using the masking slider and some of the others in the Detail panel –
And this is a look at using masks to direct different levels of noise reduction and sharpening (more sharpening and noise for the mice, less sharpening and noise for the background) –
Neither is terrible, but certainly no where near as good as the denoise AI function they just released. That’s why so many of us had to use other software to get it done. I have the standalone version of Topaz Denoise AI, so let’s run it through that without any additional sharpening in Lr –
I used the Standard model with the automatic settings and no post processing adjustments to details or color noise. Clear was too smooth looking and low light didn’t get rid of enough noise. Severe noise wiped out almost all the detail in the fur so that wasn’t good either. This worked best for sure, although compare the top dark part in this shot and in the first one – Lightroom smooths things so much nicer and the transition from really out of focus to sharp is better. Without the new tool though, Topaz produced the best result.
So what will I use now?
As I’ve played with denoise from Adobe and Topaz, I found that occasionally Topaz does a better job with preserving details. Each software seems to treat those details differently and I’ve noticed in the same photo that one will do a better job in one bit with the other doing worse, and vice versa. That seems to be pretty normal and will vary from photo to photo. And of course not all images will need this level of denoise – most of the time the old method will be just fine. No extra DNG file for that either.
Since Topaz used a perpetual license model for its separate Sharpen and Denoise AI plug ins and they work just fine, I’ll keep them ready to go. I can’t see a need for Photo AI at all with the product where it is at the moment. Using Adobe’s new tool has given me confidence in what can be achieved natively in Lightroom with occasionally moving to Photoshop for more control with Smart Sharpen, layers and masking.
My dream for the future
What I really want is to direct the amount of sharpening and/or noise reduction using selections and masks without having to go into Photoshop and create a bunch of layers. Doing that definitely produces the best results, but it’s more time consuming, makes extra steps and even more files on your hard drive. If I could tell it how much denoise to apply in different areas for a single denoise operation, that would be really helpful.
Usually dark shadows and backgrounds will need less or maybe no sharpening, but a lot of noise reduction. Conversely, subjects will need enhanced detail, but can carry more noise compared to backgrounds. Some parts of photos will need some sharpening, but not as much as the subject. And the amount of denoise required will be different as well. Right now Topaz and other products will let you mask, but as far as I can tell it’s on or off. We can’t say apply 50% noise reduction over here, but 100% over there. Nor can we do that with sharpening naively in any app using AI analysis technology. Only with image blending can we take photos made with different strengths of denoise or sharpening applied globally or with binary masks, to combine into one picture with the proper amounts all around.
I’m not sure it’s even possible given that a new file is created with AI denoising, but a girl can dream.
Last minute addition
Where Topaz Denoise AI is the clear winner –
And with Lr denoise & sharpening –
So yeah, I think there is room for improvement and Topaz will continue to earn its seat at the table.
Leave a Reply