Renormalization is a method to extract finite results comparable with the experimental values from the infinite results of QFTs. Such as the theoretical desription of Lamb shift and anomalous magnetic moment of electron. Nevertheless the method is incomprehensible physically as beyond the poineers like Heisenberg and Dirac also S. Weinberg mentioned in a footenote of his book "Qunatum theory of fields": (by analogy) "No two expert of renormalization agree"!
Theoretically renormalization is needed in view of the discrepancy between the order of differential operators (~ propagators) in QFTs which are up to two and the order of integral operators in 4D QFTs which is four.
In its QED version where renormalization the is worked out needs beyond the rgulariozation which menas geometrically a compactification of integral ranges, several dimensional conditions like Ward identity (L^2 = 0), constant magnetic fields (1/L^2 = cte.)and plane waves (2 degrees of freedom). These althogether means in my geometrical view a reduction of 4D QED to its 2D version. Then 2D QFT are by definition renormalized in view of the fact that in this case the mentioned orders of differential and integral operators are two. Also in the case of more incomprehensible renormalization of non-abelian QFTs there are similar dimensional conditions where a. o. some 1/L^2 dimensional (B-field of order B^2) is neglected (set to be zero) and the 1/L^2 dimensional Landau gauge ∂μBaμ=0 and the gauge fixing for Faddaey-Poppov ghost field ∂μCμ=0!
Considering all geometrical and dimesnional conditions on QFTs such as compactification of integration manifold by regularization/cut off and vanishing or constancy of several dimensional quantities as above renormalization menas geometrically as mentioned above the rduction of 4D QFTs to their 2D versions.
I started a geometrical program to prove this general statement by the following work:
"Topological approach to renormalization" added here.