When do we terminate the random walk? And do we have a good metric of success for use in this algorithm? Besides just how it looks to us
Since this is a local algorithm, isn't it possible for it to stuck in some bright area surrounded by a dark area? Should we restart occasionally to prevent that from happening?
what exactly do we mean by "normal" here, since we're just looking at an image, right? Not an actual model of the city?
How much longer is the long walk compared to the short walk, and can the values used in the short walk still be used? (perhaps the differences in adjacent pixels are not noticable to the human eye, but a computer can still use that data, perhaps?)
If we don't limit the maximum time steps of the long walk, could we get a picture that is pretty similar to the original image in not enormously large number of steps?
Approximately how different in time does it take comparing short and long walk?
Is it possible to start multiple short walks to simulate a single long walk?
How is the "occasionally" quantified here?
Can we recursively do short walks in parallel to build up different parts of the image?