Fig. 2.

Algorithm for DM shape optimization. (a) Algorithm overview. At each iteration, three images are taken by the camera, and PSFs in each view are measured using metric M. (b) Total weighted metric convergence (left panel), and individual-view metric convergence (middle and right panels). See Visualization 2 ^{(51.5MB, avi)} for PSF changes over iterations. Notation: $C_t$ is the DM command vector at time t, g is the gain, $D_t$ is the vector of random perturbations, $M(C_t)$ is the metric of PSF image obtained after command $C_t$ was applied to the DM. Scalebar 2 µm.