2
$\begingroup$

$\newcommand \X {\mathcal{X}}$ $\newcommand \sd {d_{\rm sign}}$ Let $(\X, d)$ be a metric space and define the distance between a point $x \in \X$ and a set $S \subset X$ by $d(x,S) = \inf_{y \in S} d(x,y)$ and the signed distance to be $$ \sd(x,S) = \begin{cases} d(x,S) & \hbox{ if $x \not \in S$}\\ -d(x,S^c) & \hbox{ if $x \in S$} \end{cases}. $$

For any $\sigma \in \mathbb{R}$, we can define $S_\sigma = \{x \in \X \,:\, \sd(x,S) \leq \sigma\}$. When $\sigma > 0$ this corresponds to enlarging the set $S$ and when $\sigma$ is negative, this corresponds to shrinking $S$. Is there a name for the set $S_\sigma$, and do these sets have well-known properties?

$\endgroup$
1
$\begingroup$

Although your context is broader, these level sets under the Euclidean metric are known as offset polygons. Here is an image from an earlier MO question/answer:


 
Another useful term in this context is the Minkowski sum, e.g., this PDF slide presentation by Andreas Bock: Minkowski Sums and Offsets of Polygons.

$\endgroup$
3
  • $\begingroup$ Awesome, thank you. I am also interested in the following operation: Fix some $\sigma > 0$ and map each set $S$ to the double-offset polygon $(S_{-\sigma})_\sigma$ where we first shrink it by a distance $\sigma$ and then enlarge it again. This has the effect of smoothing out any details of the set $S$ that are smaller than $\sigma$. Does this have a name? $\endgroup$
    – Travis
    Feb 15 '15 at 22:14
  • $\begingroup$ @Travis: Interesting smoothing operation. Surely this must have been investigated, but I am not recalling references... $\endgroup$ Feb 15 '15 at 22:46
  • $\begingroup$ @Travis: This is a stretch, but I just remembered that one can smooth a shape using the medial axis: "Shape Smoothing Using Medial Axis Properties," (journal link). And the medial axis is very much related to your $S_{-\sigma}$. $\endgroup$ Feb 16 '15 at 0:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.