SUMMARY
When i run a fuzzy search to find duplicate images, i get the best results with digikam compared to other tools. So digikam is my choice. 

But the problem is, i can not do anything with the results. I know, create a button, which saves the largest image could be easy (reverse call: delete all duplicates except the largest one). 
I cannot use the referenzce-image, because all my relevant images are in the same directory (sorted by year and month). 

I want a export-button for the fuzzy-search result. It should give a json file (which is easy to use with jq for shell-scripts) like this
{ reference-image.jpg: 
     path: "/path/to/reference-image.jpg";
     size: "5,4GB";
     pixel: "3900000";
     resolution: "2600x4500";
     duplicates: [
          { duplicate-1.jpg:
               path: "/path/to/duplicate-1.jpg";
               size: ...
           };
           { duplicate-1.jpg:
               ....
           };
     ];
}

The exported fields should be configurable like the columns in tableview from the search.

So i can check the pictures in digikam, if the result is ok, export the filenames and run my own scipt which can preserve or delete several files (the biggest one, the one with name in nameconvention, if some duplicates are the same except the filename... and so on.)

With an export like this, your are not in the situation to decide for your users, and implement a bunch of algorithms to delete duplicates which do not fit the usecases for all users... 

SOFTWARE/OS VERSIONS
Fedora 32, Gnome Desktop