为了给这个问题加上一些数字，我从 scikit-image ( skimage.feature.hog) 中实现了一个基本的梯度直方图。这是 HOG 的时序数据，默认参数应用于skimage.data.astronaut黑白图像并重新缩放以具有给定尺寸：

Image dimensions......(102, 102)
3.49 ms ± 26.3 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
Image dimensions......(256, 256)
27.5 ms ± 295 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
Image dimensions......(512, 512)
123 ms ± 2.88 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
Image dimensions......(768, 768)
284 ms ± 9.18 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
Image dimensions......(1024, 1024)
509 ms ± 8.62 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
Image dimensions......(2048, 2048)
2 s ± 15.7 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

这是np.fft.ifft2对同一组图像进行傅里叶变换的傅里叶逆变换 ( ) 的时序数据。（注意，取傅里叶变换不包括在计时结果中）

Image dimensions......(102, 102)
684 µs ± 32.2 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)
Image dimensions......(256, 256)
2.51 ms ± 26.1 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
Image dimensions......(512, 512)
21 ms ± 519 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
Image dimensions......(768, 768)
42.5 ms ± 621 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
Image dimensions......(1024, 1024)
85.9 ms ± 1.47 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
Image dimensions......(2048, 2048)
381 ms ± 9.18 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

时序数据图