Computing visual attention from scene depth | IEEE Conference Publication | IEEE Xplore