About Sentinel-1 Subswath Noise Removal Problem using FFT

I don't why I cannot to find the correct solution to achieve my goal, removing sub-swath noise in Sentinel-1 HV polarization imagery. I’ve tried as my best in StackOverflow website and CSDN in Chinese. But, those efforts are meaningless. I still don't know how to using FFT in python to get my succeed.

I've referred to these website(https://stackoverflow.com/questions/38476359/fft-on-image-with-python , https://stackoverflow.com/questions/32951036/opencv-python-mask-on-fftimage-why-do-we-need-two-channels ) and some PPT on introducing Fourier Transformation. Besides, I get five steps to make FFT, as following:

1.read image

2.get fft of image --> f

3.crate mask

4.multiply f with mask --> g

5.get inverse of g

My puzzle encountered in coding is how to put FFT apply to tiff format images, websites in above is for processing .jpg or .png format. Another question is when I found a sample demo about processing tif( Coding is given in the end of the article), I unexpectedly to find the result after FFT existed a fatal bug, data value of tiff image changed. Until now, I've wasted a week days in the begining of  new semester of my master degree in the last year. Oh, my god!

I surrender and give up for this way. Maybe I won't remove boundary noise in S-1 HV images.

Coding:

# -*- coding: utf-8 -*-

"""

Created on Tue Sep 04 14:12:09 2018

@author: Baikal

"""

import cv2

import scipy

import numpy as np

tif = cv2.imread( r'C:\Users\Baikal\Desktop\delete\Aim_Beta0_HV.tif', 2 )

rows, cols = tif.shape

## Get FFT imgage.

fft = np.fft.fft2( tif )

## Move to center

fftshift = np.fft.fftshift( fft )

## Create Mask tif file.

rowl = rows / 2 - 100

rowr = rows / 2 + 100

columnl = cols / 2 - 100

columnr = cols / 2 + 100

mask = np.ones( tif.shape, np.float32 )

for row in range( rowl, rowr ):

    for col in range( columnl, columnr ):

        mask[ row, col ] = 0

        

##  Filter.    

fftshift_mask = fftshift * mask 

## Get inverse of fft.

ifft = np.fft.ifft2( np.fft.ifftshift( fftshift_mask ) )

# To show normally , to get abs of complex number(fft).

ifft_img = np.abs( ifft ) 

# Save FFT image to disk.

scipy.misc.imsave( 'ifft.tif', ifft_img )

• Original TIFF image(Data range below 0)

• FFT result(Data range from 0 to 255)