2
$\begingroup$

I have a square S, and I want to convert it to the unit disc D.

The Riemann mapping theorem says that I can do this with a conformal bijective map. But, any such mapping will cause some distortion.

Specifically, if S contains small sub-squares, they will be mapped into sub-shapes of C that are not squares. The amount of distortion depends on the specific mapping selected, and also on the placement of the small square inside S (see, for example, this nice illustration: the distortion is minimal near the center, and maximal near the corners).

The amount of distortion can be quantified in the following way: Let $s$ be a small sub-square contained in $S$. Let $d(s)$ be its image under the conformal mapping ($d(s)$ is contained in $D$). Let $m(d(s))$ be the maximum-area square that is contained in $d(s)$. Define:

$$Distortion(d,s) = area(d(s)) / area(m(d(s))) - 1$$

So, if $s$ is mapped to a square, then $d(s)$ is a square, $m(d(s))=d(s)$, and $Distortion(d,s)=0$. On the contrary, if $s$ is a very distorted shape, probably $m(d(s))$ will be quite small, and $Distortion(d,s)$ will be large.

For each mapping $d$, it is possible to define its maximal distortion for squares of a certain size x:

$$MaxDistortion(d,x) = max[over\ all\ squares\ s\ of\ size\ x\ in\ S]{Distortion(d,s)}.$$

similarly:

$$AvgDistortion(d,x) = avg[over\ all\ squares\ s\ of\ size\ x\ in\ S]{Distortion(d,s)}.$$

My questions are:

  • For a given Riemann mapping d and square-size x, how can I find $MaxDistortion(d,x)$ and/or $AvgDistortion(d,x)$?
  • Of all Riemannian mappings from S to D, how can I find the Riemann mapping with the smallest MaxDistortion/AvgDistortion (for a given square-size x)?

If a similar question has been studied, even with different definitions of distortion, I will also be happy to know.

$\endgroup$
1

1 Answer 1

2
$\begingroup$
  1. Question: In general it seems difficult to find max and mean distortion as a function of $x$. For the square is looks doable. The following papers describe numerical code to find the Riemann mapping for arbitrary simple curves (described by landmark points).

    • Sergey Kushnarev: Teichons: Soliton-like Geodesics on Universal Teichmüller Space, Experimental Mathematics, Volume 18, Number 3 / 2009

    • E. Sharon and D. Mumford, 2D-Shape Analysis using Conformal Mapping (pdf), Proceedings IEEE Conference on Computer Vision and Pattern Recognition, Washington, DC, June 2004. (pdf)

  2. Question: The Riemann mapping is unique up to composition with an Moebius-Transformation in $SU(1,1)$, (which fixes the unit circle) $z\mapsto \frac{az + b}{cz+d}$. The effect of the Moebius transfomation on areas is easily computable.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .