Meaningful automatic video demultiplexing with unknown number of cameras, contrast changes, and motion