Scipy 1D平滑保留边缘

Question

我有一个数组，我想在不改变边值的情况下使用Scipy的gaussian_filter1d应用1d高斯滤波器：

>>> from scipy.ndimage.filters import gaussian_filter1d
>>> arr = [1.,2.,3.,4.,100.,5.,6.,7.,8.,9.,10.]

然而，这无法保留边缘值1和10：

>>> gaussian_filter1d(arr, 5, mode="nearest")
>>> array([  8.26708045,   9.54770677,  10.72077793,  11.70511557,
            12.43900171,  12.89021341,  13.06040738,  12.98285813,
            12.71434815,  12.32350223,  11.87856682])

Answer 1

这种方法似乎不必要地复杂化，但是这会使梯度平滑而不是值本身：

# Convolve with sobel filter
grad = convolve(arr, [1,-1,0], mode="nearest")[:-1]
# Smooth gradient
smooth_grad = smooth_grad = gaussian_filter1d(grad, 5)
# Integrate
smoothed = [arr[0] + sum(smooth_grad[:x]) for x in range(len(arr))]

产：

[1.0, 3.6, 6.0, 9.6, 10.7, 11.3, 11.5,11.2,10.7,10.0]