Further to defining the central point, we may also want to respect face width's/height's for the case of the face being bigger than the thumbnail. We'd need to do some pre-scaling down in the convert_image_plus function, so that what is passed through to convert_image has a face fitting the thumbnail with some spare.