NASA Vision Workbench

Government doesn't make anything easy. However VW is trying to be relicensed under Apache 2.

Honestly I don't know enough about ITK to compare it to VW. VW's original developers just wanted to make a cool way of doing lazy image evaluation. This is its primary difference compared to OpenCV. VW leverages that ability to handle really large images (1+ GB) and to do multithreaded rasterization.

Saying VW was created because of "Not Invented Here" is a little harsh. It is mostly just a fun exercise of CRTP that grew into a supporting library. Is this the best library for you? That depends. In my opinion, VW's strength at the moment is cartographic applications.

I would guess it's a combination of Not Invented Here syndrome and different problem spaces. Despite the fact that ITK seems to focus on very small things, while NASA tends to focus on the quite large, I've found that image processing, by and large, is image processing, so I'd think the former is probably more important.

Or OpenCV for that matter?