A proposal for high compression of faces in video sequences using adaptive eigenspaces