What's the advantage to using L2 norm instead of Hamming norm?

[Amoko]

the code in function ''normalized_distance(_a, _b)'':

        norm_diff = np.linalg.norm(b - a)
        norm1 = np.linalg.norm(b)
        norm2 = np.linalg.norm(a)
        return norm_diff / (norm1 + norm2)

[arnavkagrawal]

As per the paper based on which this code is implemented (An image signature for any kind of image, Wong et al):

Relative to L1 norm(Hamming distance), L2 norm gives greater emphasis to larger differences. This is appropriate because a difference of 1 at a coordinate may be due to roundoff, but a difference of 2 is meaningful.

[Amoko]

As per the paper based on which this code is implemented (An image signature for any kind of image, Wong et al):

Relative to L1 norm(Hamming distance), L2 norm gives greater emphasis to larger differences. This is appropriate because a difference of 1 at a coordinate may be due to roundoff, but a difference of 2 is meaningful.

I got it! thx bro.