Normalisation of haralick feature computation

Hi all OTB users,

I am using otbcli_HaralickTextureExtraction inside my chain, it works pretty well. And it is the only one I found, which allows to get one value for each pixel of the image associated with each haralick attribute (and not only one value by image for each attribute).

However I faced some issue when I compare the results obtain with S2 images of the same tile acquiert at different dates.

I become aware with different processing on different S2 images that the values obtained within a same image are not independent. Indeed when there is snow or cloud,… on a part of the image, it seems that a bias appears for all the values within the image.

If someone already face such issue, I would be happy to get some advices of pre and/or post processing in order to obtain comparable values over time on my area of interest whatever the spectral content of the other parts of the image.

Thanks a lot for your help
Aline Déprez

Dear @depreza,

I have never used this application, thus I don’t have particular advice to give. But I saw you opened an issue, and I will try to find what is the source of this undesired behavior.

This is a real problem, I want to tell you!