The scenario
You need to preprocess images for a computer vision model, ensuring pixel values are linear (scene-referred) before feature extraction or training.
What is in this drawing
Read the decisions behind it.
01
Proceed with RAW sensor pipeline vs standard decode
02
Apply inverse gamma or camera response curve when gamma-encoded
03
Normalize to 0-1 after linearization
Use when building pipelines that ingest camera RAW or gamma-encoded images and require linear light input for accurate model performance.
More like this