Express similarity measure in terms of probability


I have an image called 'source.png'. And a set (lets call it set-1) of images called 'target-i.png' (where i:1-10). I need to find which of the images in set-1 match 'source.png'. To acheive that, I am doing a hostogram match between source.png and every image in set-1.

The difference calculated is a Eucledian distance of the difference between the histogram of the compared images (in grey scale coverted form). Each output is a scalar value. Lets assume the output of the generated by comparing 'source.png' with every image in set-1 is as follows. [0.08, 0.015, 0.06, 0.03, 0.035, 0.045, 0.02, 0.08, 0.09, 0.01].

I assume that if the distance is less than or equal to 0.04 then I classify those pairs as matches. With this theory, we see that images target-2, target-4, target-5, target-7, target-10 are the matches for 'source.png'.

In the above approch, the matching criteria is expressed as either TRUE(1) or FALSE(0).
if compare (source.png , target-i) <= 0.04
result = 1;
result = 0;

I would like to express the result in terms of probability (%). That is with what probability (%) target-i.png matches with source.png ? Is it a 50% match or 1% match etc. Please help me out figure out a way to express it.
Please can anyone suggest how to do that?

Thanks !
Last edited: