Hi folks, I'd like to ask help about evaluating curve points as displayed in the attached image.
I'm be able to discard points if the curve is straight, comparing angles between each vector segment (left part).
I'd like to ask how could I evaluate points so that I can build a spline that passes through initial points (blue ones), and adds a number of green points where needed so that I can get a curve instead of a segmented vector.
Thanks
If you don't mind a bit of math, you can always use Newton polynomials to build a n-degree polynomial that interpolates n+1 points. I prefer them over Lagrange polynomials or using Neville's scheme simply because they take O(n^2), whereas Newton takes O(n) + some
preprocessing time.
I would post an example on how to do this, but I can't get the equation tags to show the formulas and the explanation text. It does involve building a divided difference table to get a polynomial. You can either build a set of parametric equations to evaluate at any parameter, or you can put that parameter back in the divided difference table and compute it from there (faster for sure).
My preferred method for this problem is natural cubic splines.
Thanks for posting that! I actually recently did a tweener based on cubic splines, but didn't know that article existed (so I had to do all the boring algebra).
@OP: Just to add another spline type into the mix in this thread, there's also Kochanek–Bartels splines (kb splines).
Yeah, the natural cubic spline approach is better than mine if you want the curves. This way, you get cubic Bezier curves that are C2 with each other. With Newton polynomials, you'd end up having to convert from polynomial basis to Bernstein basis and then choose a simple knot vector to get it to a NURBS.
This is a bad idea. Look at this example from the same Wikipedia page you linked.I believe Lagrange Interpolation is also a valid method.
You'll be wanting a curve fron the hermite family, eg cardinal spline, kochanek-bartels, or catmull-rom.
