在Numpy中使用meshgrid / array索引切片数组

Question

I have the following code:

    big_k = gabor((height * 2, width *2), (height, width)) #Returns a 2d-array
    r = np.arange(0, radialSlices, radialWidth)
    p = np.arange(0, angularSlices, angularWidth)
    pp, rr = np.meshgrid(p, r, sparse=False)
    z = np.sum(img * big_k[height-rr:2*height-rr, width-pp:2*width-pp])

I get this error:

    z = np.sum(img * big_k[height-rr:2*height-rr, width-pp:2*width-pp])
IndexError: invalid slice

I understand this error and why it has happened. The problem is you can't slice arrays with arrays of indices. The thing is, using meshgrid is a fabulous way to speed things up & get rid of the nested loops in my code (otherwise I would have to iterate over angularSlices * radialSlices ). Is there a way I can use meshgrid to slice big_k ?

Answer 1

You need to broadcast the index yourself, for example:

a = np.zeros((200, 300))

yy, xx = np.meshgrid([10, 40, 90], [30, 60])
hh, ww = np.meshgrid(np.arange(5), np.arange(8))

YY = yy[..., None, None] + hh[None, None, ...]
XX = xx[..., None, None] + ww[None, None, ...]

a[YY, XX] = 1

the image looks like: