Raw sensor data capturing actual scene luminance — unoptimized for display. Foundation for color grading and DCP output.
You are filming a scene with a large dynamic range—from deep shadows to a window front. The camera does not record what it looks like on your monitor, but the raw light values exactly as they hit the scene. This is the Scene-Referred Image—the mathematical representation of reality, not the aesthetic interpretation for a specific display. The values are linear, not gamma-corrected, and they remain stable across different displays and output media.
In practice, this means: Your camera's raw file (RAW or a Log codec like ProRes RAW, ARRI LogC, Sony S-Log) is already Scene-Referred. The camera has not attempted to give you a finished, display-ready image on set—that would be Display-Referred and would cost flexibility in post-production. When you are later in the color grading suite, you work with this Scene-Referred raw material. You apply Look-Up Tables (LUTs) to transform it into a Display-Referred color space—for DCP, for cinema, for streaming. This separation is essential: it allows you to optimize the same shot for different output media without damaging the original raw file.
The trick is to respect this Scene-Referred logic throughout the entire pipeline. If you are working with proxy material in the edit, proxies should follow the same logic—otherwise, a hidden color correction will creep in that you can no longer trace later. The same applies to DCP mastering: the final DCP will be Display-Referred (Rec. 709 or DCI P3), but the decisions behind it are made by you and the colorist based on the Scene-Referred raw material. This guarantees consistency across all outputs—whether monitor, cinema, or streaming platform. Scene-Referred is not sexy, but it is the reason why color grading is reproducible at all.