DeepPRIME

DeepPRIME is a proprietary deep learning–based image processing engine developed by DxO Labs for RAW photo processing. It combines demosaicing and denoising into a single neural network pipeline, aiming to reconstruct accurate color and reduce noise directly from Bayer-pattern RAW data. The technology is particularly geared toward high-ISO and underexposed images, where conventional pipelines often struggle with chroma noise, color artifacts, and loss of detail.

Technical approach and features: DeepPRIME is trained on large datasets of RAW images and uses convolutional neural networks to infer missing color channels while modeling sensor noise. The system performs joint demosaicing and denoising, preserving fine texture and edges while suppressing noise more effectively than traditional demosaicing methods. It is optimized to run on GPU hardware and is integrated into DxO’s RAW processing products, such as DxO PureRAW and DxO PhotoLab, where users can select DeepPRIME as an processing engine.

Availability: DeepPRIME is part of DxO’s commercial software ecosystem and is not open source. It is offered as an option within DxO PureRAW and DxO PhotoLab, where it is marketed as delivering higher-quality RAW conversions with improved color fidelity and detail retention relative to conventional pipelines.

Reception and considerations: Users and reviewers have noted that DeepPRIME can deliver cleaner shadows, reduced chroma noise, and finer detail, particularly at higher ISO levels. Processing times are typically longer than standard pipelines, and as with other AI-based methods, results can vary with scene content. It is generally viewed as a high-quality option for RAW conversion within DxO’s software suite.