使用Python和NumPy生成批量的n维Perlin噪声

Question

I managed to grasp the way Perlin noise works and implement a pixel-at-a-time version using this awesome PDF as a reference, but, quite obviously it's incredibly slow.

First thought would be to generate it as batches - instead going through every pixel from 0.0 to 1.0, generate a numpy.linspace and then act on it using numpy's vectorised operations, but this seems to be beyond me, I keep getting lost and deleting everything I wrote.

Could a kind StackOverflower help me? As little as example code and as much as detailed explanation of every step would both help me greatly.

EDIT: By batches I mean arrays containing values of perlin noise at different points in space, as opposed to my noise() method only generating one pixel at a time.

Answer 1

If guess correctly, then Asmagedon wants to per-calculate his noise textures and then later paint them onto some other target of a different size.

One way to do this is with scipy.signal.resample which can both interpolate and subsample a signal. From memory it only works on one dimension at a time -- just loop over the dimensions.

There are probably better ways to do antialiasing, but this one is easy.