I recommend this script in cases when masking does not solve a problem with detector gaps. The script is time and memory consuming. It can be used in cases when a nice "clear" picture is needed or with an automatic background constraction softwares (such as "Match!"). The script below uses an image data set containing gaps (<= 0 values), replaces them by interpolating from neighboring data points with a given kernel. Detailed discription is here. It saves the interpolated image to a new .tiff or .edf file (see coments in the code).

Copy the py script below in any text editor and save it in a file with a .py extension. In an .edf directory create a folder named "tiff".

You need:
numpy
fabio
glob
os
PIL
astropy.convolution

import numpy as np
import fabio
import glob
import os
from PIL import Image
from astropy.convolution import Gaussian2DKernel, interpolate_replace_nans, convolve

# image path and format
path = r'C:\scripts\image'
files = glob.glob(path + "/*.edf")

for filename in files:

    specname = os.path.split(filename)  # get filename
    specname = 'fix_%s' % specname[-1]  # set a new name to save

    pre, ext = os.path.splitext(specname)  # get file type
    ext = '.tiff'  # change to .tiff (for dioptase) comment if .edf is needed
    print(pre)

    imdat = fabio.open(filename)  # open image

    # rescale & make it float
    img = imdat.data[-1000:1000, 0:4439] * 1e0
    # make < 0 values to NaN values
    img[img <= 0] = np.nan

    # smooth with a Gaussian kernel with x_stddev and y_stddev)
    kernel = Gaussian2DKernel(x_stddev=3, y_stddev=0.1)

    # create a "fixed" image with NaNs replaced by interpolated values
    fixed_image = interpolate_replace_nans(img, kernel, convolve)
    img = Image.fromarray(fixed_image)

    # save in a folder named "/tiff/" that has to be created first
    img.save(path + '/tiff/' + pre + ext)