AI Color Correction for Underwater Photography: Topaz Photo AI vs DxO PureRAW Tutorial and Combined Workflow

The advanced post-production module of the AquaExposure course covers the complete Lightroom + AI workflow with hands-on exercises using your own photos.

Underwater photography consistently produces two problems that neither natural light nor flash fully resolves: digital noise (high ISO in depth, low ambient light) and lost sharpness (difficult focus, water movement, diffraction). These are two problems AI solves well, better than traditional correction tools, and without introducing ethically problematic manipulation.

This article is an applied tutorial. It takes two specific tools (DxO PureRAW and Topaz Photo AI), compares them on real underwater photography cases, and proposes a combined workflow with Lightroom. It complements the broader article on AI and underwater photography in 2026.

In 2024, during a training session in Brussels, a participant showed me photos from Komodo taken with a GoPro Hero 12 at 18 meters depth. Visible compression noise, grey-blue color cast, detail lost in shadow areas. He thought they were unusable. We processed them with DxO PureRAW, then refined the result in Lightroom, in fifteen minutes. The images went from forgotten archives to publishable photos. Not masterpieces. Honest, clean images. That two-step workflow is what this article details.

Before the tutorial, a critical framing on the limits of AI tools applied to underwater photography.

DxO PureRAW and Topaz Photo AI do two things well: reduce noise (the stray pixels that appear in shadow areas or at high ISO) and improve sharpness (recovering detail lost to diffraction or slight water movement).

These two tools do not do this: recover colors absorbed by water. The blue or green cast of a photo taken at 15 meters without flash is information physically absent from the file. No AI denoising tool reconstructs the red and orange wavelengths water absorbed. That correction stays in Lightroom, via white balance and HSL sliders.

An important nuance about Topaz Photo AI: on very noisy areas and homogeneous zones (sand, dark background, uniform water), the AI can generate slightly artificial textures. This is documented and visible if you compare the original file at 400% and the processed file at 400%. The AquaExposure position: apply denoising at 70-80% of maximum intensity, not at 100%, to preserve natural grain and avoid the "digital painting" look.

DxO PureRAW is a preprocessing tool. It processes your RAW file before Lightroom opens it, applying DxO optical profiles (distortion, vignetting, chromatic aberration) and AI noise correction.

Step 1: Import files

Open DxO PureRAW. Drag your RAW files in or import via the File menu. The interface shows a list with a preview for each image.

Step 2: Check camera support

DxO automatically displays whether your camera-lens combination is recognized in its database. A green badge confirms support. If your camera is not supported (frequent with GoPro and waterproof compacts), you still get generic denoising but no optical correction. Acceptable for archives, less ideal for important shots.

Step 3: Choose the denoising mode

DxO PureRAW offers three modes: DeepPRIME (maximum quality, long processing time), DeepPRIME XD (better at very high ISO, even slower), and Classic NR (fast, good results for quick review). For important underwater photos, use DeepPRIME. For a quick archive pass, Classic NR is enough.

Step 4: Choose the output format

DxO PureRAW can generate a linear DNG file (ideal for Lightroom, preserves full RAW dynamic range) or JPEG/TIFF. For underwater photography, always choose linear DNG. You keep the full color correction latitude in Lightroom afterward.

Step 5: Process

Click Process. On a recent machine, DxO PureRAW processes one photo in 15 to 45 seconds depending on mode and resolution. For a batch of 50 photos in DeepPRIME, allow 15 to 45 minutes.

Step 6: Open in Lightroom

DxO PureRAW can send processed files directly to Lightroom via the export-to-Lightroom option. You find the DNGs in your catalog with their original name followed by "(DeepPRIME)".

Topaz Photo AI combines three functions in one application: denoising, sharpness enhancement, and upscaling. For underwater photography, the first two are the most useful.

Step 1: Import files

Topaz Photo AI accepts RAW, JPEG, TIFF, and PNG files. This is its advantage over DxO PureRAW: you can process JPEG files from a waterproof compact or GoPro, not only RAW.

Step 2: Review automatic suggestions

Topaz Photo AI analyzes each image on opening and automatically proposes a processing mode. On a standard underwater photo, it typically suggests "Low Light" (low-light noise) or "Motion Blur" (motion blur). These suggestions are often relevant, but verify them on each image.

Step 3: Adjust intensity

For underwater photography, two settings deserve particular attention.

The denoising "Strength" slider: set it between 70 and 80, not at 100. At 100, Topaz can smooth coral or sand textures until they look artificial.

The "Sharpen" slider: useful for recovering sharpness lost to depth or slight water movement. Set it at 40-60 for moving subjects (fish, marine life), at 70-80 for static subjects (macro, fixed life, wrecks).

Step 4: Preview before processing

Topaz Photo AI offers a real-time 100% preview on the selected area. Always check a critical zone (fish eye, coral textures) before confirming. This is the most important step in the workflow.

Step 5: Export

Topaz Photo AI exports as JPEG, 16-bit TIFF, or DNG. To preserve maximum editing latitude in Lightroom, export as 16-bit TIFF ProPhoto RGB. The file is larger, but all color correction capability remains available.

The AquaExposure post-production workflow for AI processing runs in four steps in this order.

1. Quick cull in Lightroom

Do not process every photo. Cull first in Lightroom (flags, star ratings 1-5) and select the images worth working on. On a 7-day trip with two dives per day, you produce 200 to 600 photos. Processing 30 to 50 with DxO or Topaz is realistic.

2. AI preprocessing (DxO PureRAW or Topaz Photo AI)

Run DxO PureRAW or Topaz Photo AI on your final selection. For RAW files with DxO camera support: choose DxO. For GoPro JPEG files or cameras not supported by DxO: choose Topaz.

3. Color correction in Lightroom

Open the DNGs or TIFFs in Lightroom. Apply your standard color correction workflow: white balance (pulled warmer depending on depth and water type), recovery of red and orange tones via HSL sliders, exposure and highlight management, haze reduction (Dehaze slider). For a detailed breakdown of this step, the article on Lightroom underwater photo editing workflow covers each slider individually.

4. Check and export

Compare the processed image against the original in Lightroom's Before/After view. If AI processing introduced artefacts (plastic textures on sand, excessive smoothing of fin detail), reduce the denoising level by going back into Topaz or DxO. For concrete examples of this transformation shown step by step, the underwater photo before-and-after editing examples article illustrates several cases.

Here is how DxO PureRAW and Topaz Photo AI perform on the most common underwater photography scenarios.

Case 1: Sony A6400 RAW at ISO 3200, 20 meters, blue water DxO PureRAW (DeepPRIME) removes luminance noise without altering the sharpness of sponge textures. Topaz Photo AI delivers a close result, very slightly more aggressive on homogeneous areas. Advantage DxO on this file type.

Case 2: GoPro Hero 12 JPEG at 18 meters DxO PureRAW does not support the GoPro Hero 12 (not covered in the optical database). Topaz Photo AI processes the JPEG directly with Low Light mode. Compression noise reduction is significant. Advantage Topaz on this file type.

Case 3: Olympus TG-7 RAW in macro, ISO 400 Both tools work. DxO generates a generic profile (TG-7 not fully covered in the database). Topaz Photo AI with Sharpen at 60% recovers shrimp antenna sharpness that DxO does not touch. Near tie, slight Topaz edge on sharpness.

Case 4: iPhone 14 Pro JPEG in housing, 5 meters DxO PureRAW: not supported for iPhone JPEG. Topaz Photo AI: effective processing in Mobile AI mode. Denoising is more visible than on a true RAW, but the result is publishable.

Case 5: Nikon Z6 RAW with moderate motion blur Topaz Photo AI with Motion Blur mode partially recovers sharpness. DxO PureRAW does not address motion blur, only noise. Advantage Topaz for this specific case.

Both tools respect the AquaExposure editorial line: they reveal what the photo contains without inventing a different scene. Reducing noise and improving sharpness means letting the viewer see what you actually photographed, by reducing the technical degradation of the capture process.

The red line remains generating new elements (Generative Fill, inpainting, replacing zones). That boundary is documented in the article on generative AI and marine deepfakes for those wanting to understand the ethical implications in depth.

The article on Topaz Photo AI for underwater photography complements this tutorial with an analysis of what AI reveals versus what it invents, illustrated with commented visual cases.

In short: if you have Sony, Canon, Nikon, or Olympus RAW files, start with DxO PureRAW. If you have GoPro, iPhone, or archive JPEG files, start with Topaz Photo AI. In both cases, color correction stays in Lightroom. And in both cases, the intensity slider stays below 80%.