Adobe just released a Lightroom update with a new AI-powered feature called Denoise that produces stunning results even if Adobe is a little bit late. About a month ago, all the buzz was about DXO’s PureRAW3 with its own very powerful DeepPrime AI-based noise reduction. The comparisons I’ve seen generally show that both of these tools are excellent.
Lightroom Denoise in Action
Lightroom Denoise is part of the Enhance feature set (including Raw Details and Super Resolution) Adobe has built into Lightroom and Adobe Camera Raw. Not only does Lightroom Denoise remove noise artifacts from raw images, but it also includes Raw Details so that image detail is not smoothed out when the noise is removed. And the results do indeed show very little loss of detail with stellar noise reduction.
One of my favorite images, shown below, was captured in late October when my friend Travis and I were driving to Cuchara and noticed a group of elk in this meadow just as the sun was setting. Typically I would bracket a backlit scene like this, but I was scrambling to find a composition before the sun dropped behind the ridge. I switched to bracketing mode just after getting this shot, but the sun was too low so there were no light rays in the bracketed exposures. Because the sun was so intense, I had to underexpose then add back +3 stops exposure in Lightroom. Hence the extremely bad color and luminance noise in this image at ISO 100.
I did notice a color shift in the enhanced version (visible in the side by side comparison above) and confirmed that exactly the same settings have been applied to both images. I think this must be caused by the color noise that was so prevalent in the original raw file-note that I had to increase the Color Noise slider to +50, and there are still visible color splotches in the shadows. With this color noise removed in the enhanced version, perhaps we are now seeing a better depiction of the true colors in the image.
Enhancing Detail and Color with Lightroom Denoise
Despite the great results of Lightroom Denoise, what got my attention was something I did not even realize was a problem. Blake Rudis of f64 Academy released a video covering Denoise in which he demonstrated how severely Adobe’s standard Color Noise Reduction affects vibrant colors in certain images.
In the past, I almost always left Color Noise removal set at the default value of 25. For most images, that setting appears to effectively remove the ugly color noise artifacts of the raw file with no apparent adverse effects. But as Blake demonstrated, in certain images, even a small amount of Color Noise reduction obliterates color details in the image. Once you see it, you will start to notice it happening in a lot of your images.
After watching his demonstration, I wanted to give Denoise a try on some of my images of the interiors of cathedrals and monuments from our trip to France and Spain. When I turned off the Color Noise Reduction on this image from the Pantheon in Paris, I immediately saw the same issues Blake had described and was astounded by how much color was being sucked out of the image by the standard Color Noise removal tool. With Color Noise set to 0, brilliant colors appeared in parts of the image that were previously dull along with nasty color noise artifacts. But with even a low setting of +10 to remove the artifacts, much of the color details were lost.
The original image was an HDR merge of two raw exposures shot underexposed at ISO 400 to maximize the dynamic range of the exposures while maintaining a fast enough shutter speed to minimize blur. This technique ensured that the images captured highlight detail at the cost of increased noise caused by the faster shutter speed. A longer exposure was not possible without a tripod, and raising ISO would only reduce exposure latitude (as explained in this post) for the highlights which were very bight. Merging the exposures as an HDR reduces the noise somewhat. (You can see more of my images from the Pantheon here.)
Lightroom Denoise currently does not work with HDR DNG images, so I had to apply Denoise to the original exposures then merge the enhanced versions to create a new HDR composite image. The results are phenomenal. In fact, I actually prefer the final image with a little bit of grain added in Lightroom because the textures are too smooth after applying Denoise. (I decided not to add grain for this example just to show how good Lightroom Denoise is at removing artifacts while preserving detail.
I don’t typically deal with noise issues in my landscape images because I typically bracket exposures and utilize ETTR techniques when possible. But there are exceptions, as evidenced by the sunset landscape discussed above, including nightscapes and astrophotography along with single exposure images captured under less than optimal conditions, such as a metered exposure at ISO 400. Such images may have minimal noise, but Denoise can effectively eliminate what noise there is without otherwise degrading the image.
The final comparison shown below is one frame from a long exposure panorama of the foreground that became part of a panorama of the full arch of the Milky Way. This image is a 64-second exposure captured under starlight with a Sony a6000 APS-C camera at ISO 5000. The comparison shows just how incredible the results from Lightroom Denoise can be. For the original image, I aggressively applied the manual noise reduction tools in Lightroom to improve the image as much as possible. You can easily see that there is no question that the Denoise result is far superior. I’m thinking I may have to re-process the panorama (again).
I’ve provided a few more examples below of Denoise applied to some previously processed images. For these images, Denoise was applied at the default setting of +50.
Denoise with Infrared
Lightroom Denoise also works great for infrared images which often show some noise even with good exposure at base ISO. With mirrorless cameras, infrared images can also show horizontal lines across smooth areas of the image such as the sky. These are apparently caused by reflections off of the PDAF pixels embedded in the image sensor. I experienced this problem with my Sony a6000 infrared camera. These could not be fully removed in Lightroom, so I usually just tried to hide them with grain or else blur the sky in Photoshop for important images.
In this image, the lines were visible in the lighter part of the sky near the mountain summit but were effectively removed by Denoise. I felt like the Denoised image was too smooth for a black and white photo, so I added back a little bit of grain.
If you have not tried out Lightroom Denoise on your images yet, I highly recommend it. I am quite wary of the impacts of AI technologies on photography, especially the ability to generate completely fake images, but these kinds of AI-based tools that help improve our real images are incredibly helpful.