Investigate possible performance gains about rscolorq

I did some profiling and a lot of the time is spent in the bounds checking of the matrix. You could remove most of it by providing a get_row and get_row_mut methods and pre checking bounds of loops and then just get_unchecked

[okaneco]

I did some profiling and a lot of the time is spent in the bounds checking of the matrix. You could remove most of it by providing a get_row and get_row_mut methods

That's a good idea. I have a branch where I've added those methods to Matrix2d and rewrote some of the loops to use chunks where possible. There is a small speedup as image resolution increases. The PR is here #11.

I experimented with get_unchecked but it wasn't decisively faster to use, so I didn't push those changes and would prefer to stay with safe for now. I'm not sure where I'd have to sprinkle in asserts or bounds checks to give more hints to the compiler. While there is a lot of time spent in bounds checking, I believe the algorithm overhead is large enough to outweigh it. However, there may be ways to rewrite or optimize parts of the algorithm. More extensive profiling is needed when I get the time.