Black stripes removed but are shown again after binarization

Hi everyone,

I have read about GRD Noise Border Removal in this forum. I have used it and the result was fine, It removed all black stripes on the border. Problem is when I bimarize image, the removed back stripes show again as white stripes. It is problem becauce I want to binarize image to detect flood. Value 1 is set to removed striped so it has same value as my flood pixels and it is described as flood when I want to make vectorization in QGIS and make further processing.

After border noise removal there is no black stripe, also after calibration.

But after speckle filtering there is some black line on the previous border.

And when I binarize it and then vectorize in QGIS, this is the result. Removed black stripe is still there and has value 1 as water -> band math Sigma0_VV<value is 1, I expect the value of black stripe is 0 and after thresholding is set to 1.

When I modify band math to Sigma0_VV < 2.8E-2 and Sigma0_VV > 0, wide border stripe is removed but there still some “artefacts” on left side.

Is it possible to absolutely remove these border lines during processing ? I could do the subset but I want process lot of images automatically and every image has stripe on different side or with different size. Also every image has little bit different number of pixels so I do not know how could I automatically do subset without removing too much of data.