Résumé : Imaging Atmospheric Cherenkov Telescopes (IACTs) need imaging optics with large apertures and high image intensities to map the faint Cherenkov light emitted from cosmic ray air showers onto their image sensors. Segmented reflectors fulfill these needs, and as they are composed from mass production mirror facets they are inexpensive and lightweight. However, as the overall image is a superposition of the individual facet images, alignment is a challenge. Here we present a computer vision based star tracking alignment method, which also works for limited or changing star light visibility. Our method normalizes the mirror facet reflection intensities to become independent of the reference star's intensity or the cloud coverage. Using two CCD cameras, our method records the mirror facet orientations asynchronously of the telescope drive system, and thus makes the method easy to integrate into existing telescopes. It can be combined with remote facet actuation, but does not require one to work. Furthermore, it can reconstruct all individual mirror facet point spread functions without moving any mirror. We present alignment results on the 4 m First Geiger-mode Avalanche Cherenkov Telescope (FACT).