Back
Gigapixel AI
Sharpen AI
DeNoise AI
Video Enhance AI

Introducing the new DeNoise AI RAW model: remarkable noise reduction using RAW files

AUTHOR:
Brian Matiash
Published:
September 29, 2021
Time to read:
10 Minutes

Remarkable image noise reduction for your RAW files

We’re SO excited about our newest model in DeNoise AI v.3.3, built specifically to take advantage of the vast amount of image data in your RAW files. We use whatever image data you send to DeNoise AI, so you can imagine how much cleaner your noise reduction results could be when you send RAW sensor data as opposed to processed RGB data (like you’d find in a JPEG file).Our RAW model uses all of that rich sensor data to provide results that are cleaner than anything else out there, even our existing models! We also made serious improvements to our DNG output support, so you’ll still be able to edit your saved image files with the same precision as your RAW files. Here are some example photos that showcase just how powerful the new RAW model is.

Original Photo | © Hillary Fox
Noise reduction using DeNoise AI RAW model | © Hillary Fox
Original photo | © Hillary Fox
Noise reduction using DeNoise AI RAW model | © Hillary Fox
Original photo | © Hillary Fox
Noise reduction using DeNoise AI RAW model | © Hillary Fox

Comparing the DeNoise AI RAW model to other models

It’s one thing to show you some compelling before and after photos. It’s also important to visualize what I mean when I say that the RAW model produces results that are cleaner than anything else out there. Here’s a Comparison View of that parrot photo I shared above. At first glance, it may be difficult to see the differences between the RAW model and three of our other models.[

Now, let’s apply a Curves adjustment and desaturate this same Comparison View photo. This allows us to see just how much better the RAW model is at eliminating that distracting mottling.

Comparing the DeNoise AI RAW model to Lightroom Classic

We also wanted to compare the results of our RAW model against Lightroom Classic’s noise reduction. Not only is there a lot of mottling, but Lightroom also obliterated so much important edge detail all over the parrot’s head.

How does the RAW model work?

The best way to explain the benefits of the new RAW model is to compare how it works with our four existing DeNoise AI models: Standard, Clear, Low Light, and Severe Noise. Let's say you want to use DeNoise AI to apply noise reduction to a RAW photo you have stored in your Adobe Lightroom Classic catalog. Currently, the supported workflow requires you to convert your RAW file to a processed RGB file format that DeNoise AI uses as the input source.

Sourcing RAW image data vs. RGB image data

Unfortunately, RAW converters introduce clipping, demosaicing, and other post-processing steps when converting the RAW data to an RGB image. Different RAW converters use varying conversion methods that change the noise pattern of the sensor data. When we train a model for RGB images, we take the variations of these RAW converters into account. Also, the RGB images can go through additional editing processes by the users, which we also account for in our RGB model training.As a result, a great deal of effort is required to cover all possible variances when using RGB models to apply noise reduction. We can eliminate this extra effort with our new RAW model by utilizing the pure RAW data from the camera, where noise is ideal. The RAW model can now learn the unchanged noise patterns and apply better noise reduction without the inherent flaws of processed RGB files.

Using Bayer filter data for improved results

The RAW model also does the demosaicing itself, which should be much better than traditional demosaicing algorithms because we use the original Bayer filter data that the camera sensor records as the input source. The primary benefit of using the Bayer filter data as a file input is the number of color channels, and most cameras use it when saving RAW files. The Bayer filter pattern contains four channels for each pixel on the camera sensor: two green, one red, and one blue. Typically, a demosaicing algorithm is then applied to interpolate a red, green, or blue value for each pixel, resulting in a 3-channel processed image file.

Using the Bayer filter data allows DeNoise AI to produce better image detail while minimizing residual patterns. The result is an output image that is more detailed with fewer artifacts and aliasing. In other words, we designed our new RAW model to take full advantage of your unprocessed RAW files, allowing us to deliver superior noise reduction, especially when compared with using JPEG files. We also have this great article that dives into the differences between RAW and JPEG files.

Let's talk about workflows.

We've extensively discussed photo editing workflows on the Learning Center, and it is, after all, one of the most popular topics that we keep hearing. The new DeNoise AI RAW model is the next step to provide outstanding noise reduction results without sacrificing image quality, but we're still working to improve workflow functionality.

When should I send my RAW files to DeNoise AI?

This is critical: You need to send your RAW file to DeNoise AI at the very beginning of your editing workflow to get the most benefit from the RAW model. That doesn't mean that you have to send every RAW file to DeNoise AI. You're encouraged to go through your current rating and culling processes to identify the images you're most likely to edit.

Which RAW files are supported?

Before diving into workflows, it is important to understand that the RAW model will not convert a non-RAW file (JPEG, TIFF, PNG) into a RAW file. The RAW model will only work with supported RAW files. While we support most RAW file formats from major camera manufacturers as input files, the RAW model does not currently support DNG or Fujifilm RAW files. Part of these limitations involves how we currently handle DNG files and how Fujifilm X-Trans sensors differ from more common Bayer filter CFA sensors. The RAW model will also be greyed out if you open a processed RGB file (JPEG, PNG, TIFF) or a DNG file.

DeNoise AI standalone workflow

This one is pretty straightforward. All you have to do is launch DeNoise AI on your Mac or Windows computer and navigate to the folder containing the RAW file(s) you want to edit. When you’re happy with the results, I recommend selecting DNG as the file format and Source as the destination. Doing so will save the DNG file in the same folder as the original RAW file. You can also see this workflow in action by watching the video below.

Lightroom Classic on Mac workflow

If you're an Adobe Lightroom Classic user on the Mac, there is an option to drag and drop a RAW file from the Grid view onto the DeNoise AI app icon to load it in standalone mode. I recommend first disabling the "Apply RAW color correction" and "Apply auto lens correction" options in the DeNoise AI Preferences window if this is your intended workflow. I also recorded this video illustrating the round-trip workflow I use to send one of my RAW files from Lightroom Classic to DeNoise AI.

Lightroom Classic on Windows workflow

It's important to note that the drag-and-drop workflow between Lightroom Classic and DeNoise AI on the Mac is not supported on Windows, and it is not something we have control over. However, you can still leverage the RAW model while using Lightroom Classic with the following steps:

Step 1. While in the Lightroom Classic Grid view, right-click on a RAW file that you want to open in DeNoise AI and select "Show in Explorer"

Step 2. Drag the highlighted file to DeNoise AI to open it. Apply the RAW model and make any additional settings changes. Then, click on the Save Image button. Be sure to use the DNG file format and ensure to select Source for the destination.

Step 3. Return to Lightroom Classic, right-click on the folder containing your original RAW file, and select "Synchronize Folder...".

We've made great strides to improve the quality of our RAW support and DNG output. However, we also understand that the current workflow recommendations aren't as smooth as we'd prefer. We are constantly working to streamline how we interact with host applications like Lightroom Classic and improve the quality of our DNG output files. For now, we hope you take the time to explore the capabilities of the new RAW model in DeNoise AI.

Try DeNoise AI for free

Visit our product page for DeNoise AI and click on the "Try for Free" button to download free trials that never expire.

Tags
No items found.
AUTHOR:
Brian Matiash

Brian Matiash serves as Product Marketing Manager for Topaz Labs and manages the Topaz Labs Learning Center. He is also a photo educator and author, with his work being featured in dozens of international publications.