I am attempting to use scikit-image PolynomialTransform to map a bunch of points to their new coordinates. Here is a minimum example

```
A = np.random.normal(0, 1, 10)
B = np.random.normal(0, 1, 10)
pt = skt.PolynomialTransform(np.array([A, B]))
src = np.random.uniform(0, 1, (65536, 2))
dst = np.zeros(src.shape)
pt.estimate(src, dst)
```

It crashes with

`Unable to allocate 128. GiB for an array with shape (131072, 131072) and data type float64`

Why does the estimate function have QUADRATIC memory complexity with respect to the number of points to be mapped ???