I really like the example! I'm not sure about the use of plt.subplots, though. I used it in a previous PR, and someone suggested that I use the older style for backward compatibility. Personally, I'm fine with it, but we should probably talk about version requirements of dependencies on the mailing list. For example, numpy version came up in the dtype._convert pull request (b/c of np.promote_types).

Thanks, fixed!

In my mind, subplots alone is worth the matplotlib 1.0 dependency. If people don't like that, we may just want to copy subplots over into skimage for now. It makes the creation of programmatic plots so much neater!

Agreed: subplots is great. Thanks for the PR. I'm merging into my branch, but I suggest we wait a day or so before merging into skimage to see if anyone else has comments about dependency versions.