Revolutionizing RAW Processing with iOS 27

Apple has announced a transformative update to its system-level RAW image processing engine with the release of iOS 27, iPadOS 27, and macOS 27. Dubbed RAW 9, this latest iteration leverages machine learning to significantly enhance detail and reduce noise in RAW photos. The update is particularly notable for its ability to reprocess older RAW images, breathing new life into existing libraries. With support for nearly 800 camera models and a new CoreML-based pipeline, RAW 9 represents the most substantial improvement to Apple's RAW processing since its inception.

What is RAW and Why It Matters

RAW is a file format that captures all data directly from a camera's image sensor, offering photographers maximum flexibility during post-processing. Unlike compressed formats like JPEG, which discard information to save space, RAW files retain raw sensor data including exposure, white balance, and color information. This allows editors to adjust settings without degrading image quality. Apple's RAW engine has evolved through nine major versions, each improving demosaicing (interpreting the sensor's color filter array), denoising, and tonal adjustments. RAW 9 takes this further by employing a tiled CoreML model that combines demosaic and denoise into a single optimized step, resulting in unprecedented image quality.

Technical Breakthroughs in RAW 9

The core of RAW 9 is a machine learning model designed specifically for RAW image reconstruction. As explained by Apple's Core Image engineers during the WWDC26 session, the model is built atop a tiled CoreML architecture, enabling it to process large images efficiently while maintaining high accuracy. The model runs entirely on-device using the Apple Neural Engine (ANE) cores, which ensures fast performance without compromising privacy or requiring a cloud connection. This on-device approach also means that all processing remains local, a key advantage for professional photographers who handle sensitive or confidential work.

One of the most impressive aspects of RAW 9 is its handling of high-noise environments. In a demonstration comparing RAW 8 and RAW 9 on a Canon 5D Mark III image shot at ISO 51,200, the results were stark. The original RAW data was virtually unusable due to extreme luma and chroma noise, obscuring the true colors of a box of crayons. While RAW 8 produced an acceptable recovery, RAW 9 delivered significantly better color accuracy and definition, even revealing shiny specular highlights on the crayons. This level of improvement is attributed to the ML model's ability to distinguish noise from actual image details, a challenge that traditional algorithms often struggle with.

Non-Traditional Sensor Patterns: A New Challenge Conquered

Another area where RAW 9 shines is with non-Bayer sensor patterns, such as those used in Fujifilm X-Trans cameras. These sensors employ a unique color filter array that is notoriously difficult to demosaic, often leading to color artifacts and loss of fine detail. In a test using a Fujifilm X-T5 at ISO 12,800, RAW 8 exhibited some artifacts and reduced clarity in embroidery yarn. RAW 9, however, produced a much cleaner result: small text became more legible, and the texture of the yarn was far more defined. This success is due to the ML model's training on a wide variety of sensor patterns, allowing it to adapt to different mosaics without relying on fixed interpolation rules.

Enhanced Detail in Low-Noise Conditions

RAW 9 isn't just about high-ISO performance; it also refines images captured in low-noise scenarios. A comparison using a Sony Alpha 7 II photo of a vintage dial indicator showed that while RAW 8 already looked good, RAW 9 produced an even sharper, clearer image with fine text that was easier to read. This improvement comes from the model's ability to reconstruct edges and subtle textures more accurately, reducing the softening that often accompanies traditional denoising algorithms.

Historical Context: The Evolution of Apple's RAW Engine

Apple introduced RAW support in macOS X 10.4 Tiger in 2005, providing basic decoding for a limited set of cameras. Over the years, the company steadily expanded its compatibility list while refining the processing engine. Each major update brought improvements: RAW 1 focused on basic support, RAW 2 added better white balance, RAW 3 introduced initial demosaic improvements, RAW 4 enhanced color fidelity, RAW 5 incorporated lens corrections, RAW 6 added support for high-bit-depth displays, RAW 7 improved noise reduction, and RAW 8 brought the first integration of machine learning for select aspects. RAW 9, however, is the first to fully integrate a CoreML model for the entire demosaic and denoise pipeline, marking a generational leap forward.

The expansion of camera support has also been impressive. Starting with just a handful of DSLRs, Apple now supports nearly 800 different camera models from Canon, Nikon, Sony, Fujifilm, Panasonic, Olympus, and others. This compatibility list is updated regularly, and RAW 9 will automatically benefit existing supported cameras. Photographers using unsupported cameras can still use third-party RAW converters, but those within Apple's ecosystem will enjoy the tight integration and performance benefits of the native engine.

Developer Integration and Workflow Considerations

For developers, RAW 9 is accessible through the Core Image framework, which exposes the full RAW processing pipeline to apps. This allows photo editing software, camera applications, and even game developers to take advantage of the new engine. Apple has provided detailed documentation and a WWDC session that covers how to enable RAW 9, optimize performance for real-time editing versus batch export, and handle the new processing options. One key aspect is that RAW 9 is designed to be backward-compatible: older RAW files processed with previous versions can be reprocessed using the new engine, often yielding superior results with no additional effort from the user.

Performance optimizations were also a major focus. Because the ML model runs on the Neural Engine, it offloads work from the GPU and CPU, leading to faster processing and lower power consumption. For photographers editing on laptops, this means longer battery life during shoots. The tiled approach also allows the model to handle large sensor files (up to 100 megapixels or more) without running out of memory, making it practical for high-resolution cameras.

Implications for Professional Photography

RAW 9 is a significant tool for professional photographers who rely on precise image quality. The ability to recover detail from extremely noisy shots widens the envelope of usable ISO settings, allowing photographers to shoot in lower light without sacrificing clarity. The improved demosaicing also benefits landscape and studio photographers who need maximum sharpness and color accuracy. Additionally, the on-device processing means that results are consistent across all Apple devices running the latest operating systems, from iPhone and iPad to MacBook and iMac.

However, RAW 9 is not just for professionals. Enthusiasts using apps like Apple's own Photos or third-party editors will see immediate improvements when opening RAW files. The new engine can also process RAW images captured with an iPhone using third-party camera apps, bridging the gap between mobile and dedicated camera workflows. As computational photography continues to evolve, Apple's investment in machine learning for RAW processing positions it as a leader in camera image quality.

Availability and Future Outlook

RAW 9 is part of the iOS 27, iPadOS 27, and macOS 27 updates, which are currently available in developer beta and will ship to the public later this year. Apple has not yet announced specific hardware requirements, but given its reliance on the Neural Engine, devices with A12 Bionic chips or newer are likely required for full functionality. Older devices may fall back to RAW 8 processing.

Looking ahead, RAW 9 sets the stage for even more advanced processing techniques. As CoreML models become more sophisticated, future updates could incorporate scene-specific optimizations, such as better handling of astrophotography or underwater photography. Apple has also hinted at expanding the camera support list to include even more niche camera models. For now, RAW 9 represents a massive step forward, proving that the combination of machine learning and traditional image processing can achieve results that were previously unattainable.

Source:9to5Mac News