At one point I did the same for a comic app which was getting the earliest releases of a manga I wanted to read; I still don't read Japanese but was the buyer for my translation circle. They had similar forms of obscure obfuscation; They scrambled the image into chunks, then you got a metadata that remapped it into a finished image. Raw, it looked like one of those slide puzzles.

Over the course of a couple years they updated their scrambling; First to randomize the size of the regions, then to make them triangular instead of rectangular. It was an interesting if tedious challenge to reverse engineer.