You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi @Dobatymo
I believe that both variants are equally common in the literature. Wikipedia article has the summation outside, while other sources (see image) put it inside. We have exact formula included in the docs so user can decide if it satisfies his use case or not.
Feel free to close the issues if it answers your question!
Hi @zakajd Sorry I missed the formula in the docs. However both Wikipedia and the two references from the docs have the sum outside. I am not familiar with any formulation which has the sum inside. I am only familiar with the isotropic and anisotropic formulations. However both have the sum outside (well it only matters for the isotropic version). Only the sum of the per pixel norm differs.
The total variation (l2 version) is calculated here as
sqrt(sum(d_w**2 + d_h**2))
. Shouldn't it besum(sqrt(d_w**2 + d_h**2))
instead? Seepiq/piq/tv.py
Lines 34 to 37 in 26d044e
Now the problem is how to vectorize this correctly...
The text was updated successfully, but these errors were encountered: