Adaptive Convolutions

TL;DR

This paper develops a theoretical framework for adaptive convolutions that locally adjust smoothing based on the function's variation, using a matrix-valued adaptation function derived from phase space transformations.

Contribution

It introduces a formal theory for adaptive convolutions and derives a formula for automatically selecting the adaptation function based on local variation of the function.

Findings

Provides a mathematical framework for adaptive convolutions.

Derives a formula for the adaptation function $$ based on local variation.

Ensures invariance under shifting and scaling of the function.

Abstract

When smoothing a function $f$ via convolution with some kernel, it is often desirable to adapt the amount of smoothing locally to the variation of $f$. For this purpose, the constant smoothing coefficient of regular convolutions needs to be replaced by an adaptation function $\mu$. 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 $\mu = \mu_f$ in dependence of the function $f$ to be smoothed. This requires the notion of the \emph{local variation} of $f$, the quantification of which relies on certain phase space transformations of $f$. The derivation is guided by meaningful axioms which, among other things, guarantee invariance of adaptive convolutions under shifting and scaling of $f$.