A method extracts an alpha matte from a video acquired of a scene. A set of pinhole or narrow aperture images I.sub.P is acquired of the scene with a camera aperture set to a relatively large depth-of-field. The scene includes a background B and a foreground F. A corresponding set of wide aperture images I.sub.F is acquired of the scene with the camera aperture set to a relatively small depth-of-field. The respective pinhole and wide aperture images are combined to extract an alpha matte according to .alpha.=1+(I.sub.F-I.sub.P)/(B-B.sub.F).