image processing methods

[v923z]

Taking ideas from #363, this discussion should explore, what kind of image processing methods could be useful in the context of a microcontroller. A good first candidate could be openmv's projective transforms, as well as blob searching.

@kwagyeman
@teuler

[kwagyeman]

@v923z - OpenMV maintains a library of image-processing functions. I don't have any interest in methods outside of this. I can implement and export homogrphy support into our code base directly. However, I thought it would make more sense to have it in ulab. That said, homography is a computer vision like function... so, maybe it doesn't belong in ulab anyway? It's matrix math, though.

If svd support is exported in ulab then I can add it to openmv firmware directly.

[v923z]

When I started out with this project, I didn't name it unumpy or anything like that, because I thought that there could be functionality that is not in numpy or scipy, but might, in a way, fit into the scope of what I envisioned. And such functions are exactly the ones that I had in mind.

So, I wouldn't mind adapting functions that might have traction in the community, for the simple reason that by moving themppppppppppop would make them accessible to a bigger audience. On the other hand, I don't want to re-implementpppi@ everything here just for kicks. If no-one is interested in blobl searching, then there is no point in having a function for that.
Op
SVD was requested here in a different context, so I think it was the right decision to implement the function here.ll

[kwagyeman]

Once you start adding vision functions, people will want support, and then you to make them faster. :)

I think the homography stuff is useful, though, for non-vision functions, too. It gives you the matrix you need to map a plan to 4 points.