Glossary Term
Semantic Segmentation
AI technique that labels each pixel by object type — sky, skin, foliage — to guide colorization.
Semantic segmentation divides an image into labeled regions at the pixel level — identifying areas as sky, skin, clothing, vegetation, water, or architecture. In AI colorization pipelines, segmentation helps models apply contextually appropriate colors: faces receive natural skin tones, trees receive green hues, and skies receive blue gradients. Without segmentation awareness, colorizers may assign arbitrary colors to regions. Modern end-to-end colorization models internalize segmentation implicitly, but explicit segmentation remains a key component in research-grade restoration systems.