In ways that ensure that they are not visually recognizable on the physical medium afterwards, because the visual layout represents a whole lot of redundancy, and the job of compression is to remove redundancy. If the end result has any recognizable patterns, the compression is not doing its job well.
If you watch the video, you can see that the images are burnt to the medium and can be seen optically. With 1 and 0, you cannot do this. All you'd see would be random dots everywhere.
How are the images encoded?