Adaptive Convolutions
Abstract
When smoothing a function via convolution with some kernel, it is often desirable to adapt the amount of smoothing locally to the variation of . For this purpose, the constant smoothing coefficient of regular convolutions needs to be replaced by an adaptation function . This function is matrix-valued which allows for different degrees of smoothing in different directions. The aim of this paper is twofold. The first is to provide a theoretical framework for such adaptive convolutions. The second purpose is to derive a formula for the automatic choice of the adaptation function in dependence of the function to be smoothed. This requires the notion of the \emph{local variation} of , the quantification of which relies on certain phase space transformations of . The derivation is guided by meaningful axioms which, among other things, guarantee invariance of adaptive convolutions under shifting and scaling of .
Cite
@article{arxiv.1805.00703,
title = {Adaptive Convolutions},
author = {Ilja Klebanov},
journal= {arXiv preprint arXiv:1805.00703},
year = {2018}
}