The Mumford-Shah functional is one of the most used tools for image segmentation, but it also gives a very beautiful free boundary problem. Let be a nice domain in the plane (the screen) and g be a bounded measurable function on
. The functional is given by
Minimizers for J are known to exist, and various regularity results for K have been proved, but the conjecture that (after trivial reduction)K is a finite union of C1-curves is still unproved. I will try todiscuss recent progress by A. Bonnet and J.-C. Léger (using blow-ups andthe notion of global minimizers on the plane).