将 Boxcar 平均值应用于地理空间图像

Apply Boxcar average to geospatial image

假设以下数组 A 是读取 GeoTIFF 图像的结果,例如使用 rasterio,其中 nodata 值为 masked,即数组 B.

我想对方形邻域应用 boxcar 平均平滑。第一个问题是我不确定哪个 scipy 函数代表 boxcar 平均值?

我认为可能是 ndimage.uniform_filter。但是,与 scipy.signal 相比,ndimage 不适用于掩码数组。

from scipy.signal import medfilt
from scipy.ndimage import uniform_filter
import numpy as np

A = np.array([[-9999, -9999, -9999, -9999, -9999, -9999, -9999, -9999],
    [-9999, -9999, -9999, -9999, -9999, -9999, -9999, -9999],
    [-9999, -9999, -9999, -9999, -9999, -9999, -9999, -9999],
    [-9999, -9999, -9999, 0, 300, 400, 200, -9999],
    [-9999, -9999, -9999, -9999, 200, 0, 400, -9999],
    [-9999, -9999, -9999, 300, 0, 0, -9999, -9999],
    [-9999, -9999, -9999, 300, 0, -9999, -9999, -9999],
    [-9999, -9999, -9999, -9999, -9999, -9999, -9999, -9999]])

B = np.ma.masked_array(A, mask=(A == -9999))
print(B)


filtered = medfilt(B, 3).astype('int')
result = np.ma.masked_array(filtered, mask=(filtered == -9999))
print(result)

boxcar = ndimage.uniform_filter(B)
print(boxcar)

那么,我如何应用考虑无数据值的 boxcar 平均值,例如 scipy.signal.medfilt

这似乎是一个很好的解决方案:

import numpy as np
from scipy.signal import fftconvolve

def boxcar(A, nodata, window_size=3):

    mask = (A==nodata)
    K = np.ones((window_size, window_size),dtype=int)

    out = np.round(fftconvolve(np.where(mask,0,A), K, mode="same")/fftconvolve(~mask,K, mode="same"), 2)
    out[mask] = nodata

    return np.ma.masked_array(out, mask=(out == nodata))

A = np.array([[100, 100, 100, 100, 100, 100, 100, 100],
              [100, 100, 100, 100, 100, 100, 100, 100],
              [100, 100, 100, 100, 100, 100, 100, 100],
              [100, 100, 100, 100, 1  , 0  , 1  , 100],
              [100, 100, 100, 1  , 0  , 1  , 0  , 100],
              [100, 100, 100, 0  , 1  , 0  , 1  , 100],
              [100, 100, 100, 100, 100, 100, 100, 100]])

print(boxcar(A, 100))

希望能得到一些反馈,尤其是关于改进方面的反馈!