Still used extensively for background-forground segmentation in e.g. Photoshop, I am guessing.

Research these days mostly focuses on deep learning though.