This command has been made to be able to edit float-valued images (typically resulting from FFT) in software that does only manage 8bit-integer encoded image (GIMP 2.8).
What it does is encode each input float value as a 24bits precision decimal number, i.e as a 3x8 bits per pixel value. The resulting image is only integer valued and looks like what you mention, i.e. 3 copies of the same kind of image information along the vertical axis.
You can then save it as a .png file for instance, and reconstruct the original float pixels using
I don’t think these two commands are really useful when working from the command line, as the CLI interface manages float-valued images out-of-the-box, and can load/save them with appropriate formats (
.tiff being a good one).