Imagine that there is a famous book. Imagine that someone discovered that if you:
- Take the first letter of every sentence
- Swap the positions of every 2 letters
- Sort the middle 10 characters in the following order 3, 7, 2, 5, 9, 6, 10, 4, 1, 5, 8
- Convert the letters to a binary representation using ASCII encoding
- Perform an XOR of every binary digit with the binary exactly 9 positions to the right
- Convert the resulting data into a bitmap encoding with 100 rows
Then the result can be opened with an image viewing program, and some objectionable image will result.
Since a process of manipulating the book data can result in an objectionable image, should the book be banned?
If you feel that it should, then we should probably ban every book that has ever existed. Because, given a small enough image and any book of reasonable length, it is possible to invent an algorithm that will convert the text of that book into that image. The algorithm will be specific to that one book and that image, but the same is true of each of the images in the blockchain. The algorithm for converting a subset of blockchain data into an image is specific to the bitcoin blockchain and the reported image.
I've seen this argument before, but frankly I find it a little specious - it ignores the distinction between the ciphertext and the key.
The rule you give as an example has... maybe 40 or 50 bits of entropy. That's not enough to store an "objectionable" image, or even one image out of a giant collection of such images; we're talking about 10KB, 20KB images here, even at maximum compression. Thus, the actual carrier mechanism for the data must have been (at least in a large part) the book itself, and so holding the book responsible for distributing the content makes sense.
In contrast, a reverse-engineered algorithm like you propose is going to be much more complex, because info-theoretically, the book is no better than a PRNG that's skewed in a way you can't control. Even the length of the book doesn't matter for more than 20 bits or so (War and Peace is "only" 3.2MB, so "turn to page 100" only gives you 22 bits of entropy). So instead, you need enough moving parts in the "instructions to decode" to carry the signal; those instructions must - as a matter of pure math - be several thousand words long if written in plain English. At which point, it should be clear that it's the instructions - not the book they reference - which carry the objectionable signal.