Humans are awesome at recognizing visual patterns. It's difficult to get a sense of what a file contains from a hex editor, but if you apply the right transformations on the data, distinct features stand out in the visualization.
This allows people to recognize similarities between different files as the resulting patterns are revealed. Take this previous blog post by the same author:
This shows various sort methods operating on a set of unsorted data. If all you had visibility into was how the data was manipulated, by generating an image based on the data transformations, your pattern matching brain will quickly be able to see which sorting algorithm was used.
