One problem with all these maze algorithms is how they are completely random.
When a human creates a maze, he might start with the solution, then add a few false paths that end near the finish but just fail at the last step. Maybe a few loops are added, or a picture or special features are integrated. Only at the very end, the leftovers are filled in with random data. The process is a design, not a random generation.
Are there any algorithms available that do similar things?