Shop
My account
☰

INTRODUCING

DxO AI Masks

Taneli Lahtinen

The new standard for

masking and local adjustments

Why masking
had to change

For passionate photographers, efficiency and precision matter. 
With new DxO AI Masks combined with the power of U Point™ technology, 
DxO PhotoLab9 is pushing the boundaries of what's possible and delivering unmatched accuracy at full resolution.

The resolution
challenge

Petr Bambousek
Dalmatian Pelican, Greece
Olympus E-M1X
12-100mm f/4
ISO 400 - 1/400 - f/8 - 20mm
DxO PhotoLab

A full resolution image, here at 20 megapixels.

Here's what sets photography apart from video editing or smartphone processing: Passionate photographers routinely work with 20, 30, 
40-megapixel images, and beyond. While impressive 
AI demos showcase instant subject selection on social media, they typically operate 
at resolutions around 256 pixels in the longest dimension. That's fine for video work but it's nowhere near sufficient for serious photography.

The challenge isn't just about scaling up a small mask.
When you're working with a 50-megapixel RAW file, every edge matters,

every strand of hair, and every subtle transition between in-focus and out-of-focus areas.
Simple interpolation doesn't cut it; it leaves jagged edges, misses fine details,
and fails to capture the nuanced boundaries that define professional-quality selections.

The sophisticated five-stage

processing pipeline behind
DxO AI Masks

Our engineering team developed a cascading, five-stage processing pipeline that combines the best of modern AI with proven image processing techniques:

An example using two methods — the Subject Type “Animals”, and then selections using user-drawn bounding boxes for the reflections.

STAGE 1 :
Intelligent subject recognition (semantic boxing)

The process begins with advanced neural networks that understand what you're looking for. When you select "Sky," "People," or any subject type, the system identifies relevant objects and generates initial bounding boxes — the rough areas where your subjects exist.

Lower resolution for the first stage.

STAGE 2 :
Initial AI segmentation

Within each bounding box, we apply state-of-the-art segmentation technology to create initial masks. These operate at lower resolutions but with remarkable semantic understanding, distinguishing subjects from backgrounds even when colors and tones are similar.

The upscaled resolution at Stage 3.

STAGE 3 :
Probabilistic upscaling
(image matting)

This is where things get interesting. We employ specialized image matting algorithms — traditionally used in video production for green-screen effects — to quadruple the resolution while adding crucial grayscale information. This transforms simple binary masks into probability maps that understand partial transparency and edge softness.

STAGE 4 :
High-resolution refinement (guided filtering)

Advanced signal processing techniques, including guided filtering, bring masks to half the final resolution. This stage preserves fine details while maintaining computational efficiency — crucial for responsive editing.

Full resolution at Stage 5.

STAGE 5 :
Final Resolution

The last step scales to your full image resolution, ensuring pixel-perfect accuracy whether you're working with 20 megapixels or 100.

Beyond technology:
The integration advantage

But here's what truly sets DxO apart:
we didn't abandon what already worked. Trusted by passionate photographers for more than 20 years, our U Point technology remains fully integrated and complementary 
to the semantic DxO AI Mask technology.

Why does this matter? Because AI isn't perfect. When selecting a person against a leather chair of similar color, semantic selection excels where color-based selection struggles. But when you need to adjust only the shadowed side of a face while preserving highlights, U Point delivers precision that pure AI cannot match.

This isn't about choosing between technologies —

it's about having both at your fingertips, seamlessly integrated. Start with DxO AI Masks for complex subjects, then refine with U Point for ultimate precision. It's the workflow passionate photographers have been waiting for.

Example of integration using

a combination of Gradient and DxO AI Masks

Vitolda Klein, Jmagerie, A.C.

Example of integration using
a combination of DxO AI Masks and
U Point technology (Control Points)

Adolfo Felix

The future is here

The result? Masking technology that surpasses anything you’ve seen before. More accurate edges. Better handling of fine details. Seamless integration with existing tools. And all at the full resolution your images deserve.



Welcome to the next generation of local adjustments. Welcome to 
DxO PhotoLab9.

NO LIMITS, NO COMMITMENT

GET YOUR FREE TRIAL NOW!

Download DxO PhotoLab 9 today.

30 day free trial, no restrictions, no card details required.

*Your email address is only used for information about DxO and Nik Collection software.

Learn more about your rights and how we manage your data.

For the passionate photographer.

Learn more about DxO