math.MG Metric Geometry
- Discrete sphere packing solutions lead to error-correcting codes.
- The earthmover metric is used in image recognition and classification.
The Banach fixed point theorem for contraction mappings has a beautiful application in image compression, called fractal compression. One starts with a complete metric space $X$ of images with Hausdorff metric. Then for a given image $x \in X$ one finds a contraction mapping $A: X \to X$ with (unique) fixed point $x$. To do this, one considers self-similarities in the picture (that's why it is called fractal compression).
Then we get rid of the original image and store the map $A$ only. To reconstruct the image, one starts with any $x_0 \in X$ (for example an image which is all black or all white), and applies $A$ several times. The result will be close to $x$.
When I (Evgeny Shinder) first learnt this in high school (my friend and I implemented fractal compression as a final project for a programming class), I was fascinated how such abstract math can be applied to such a concrete problem as image compression!