What is Image Matting?
TL;DR
Technology that estimates how much each pixel belongs to the subject (its transparency) to cleanly cut out fine edges like hair. It's the core of AI background removal.
Image Matting: Definition & Explanation
Image matting estimates, for each pixel, how much it belongs to the foreground (subject) versus the background—its transparency (alpha value)—to separate the subject from the background. Unlike segmentation, which paints each pixel as a binary subject-or-background choice, matting handles continuous transparency from 0 to 1, so it can naturally cut out ambiguous edges like hair, fur, smoke, and translucent glass. Deep learning has recently enabled high-accuracy matting without a green screen, making it the core technology of AI background-removal tools. Its application to video (video matting) is advancing too, used for virtual backgrounds in streaming and for visual effects (VFX). The reason AI background removers can cleanly cut out even hair is this matting technology.