Photon-starved imaging through turbulence at the diffraction limit

Achieving diffraction-limited resolution through atmospheric turbulence is a fundamental challenge for ground-based imaging systems, especially in the photon-starved regime. Conventional methods, such as adaptive optics and speckle imaging, struggle to achieve optimal performance when turbulence is severe and photons are limited. We present Turbulence-Aware Poisson Blind Deconvolution (TAP-BD), a framework designed for robust image recovery in these extreme conditions. TAP-BD exploits phase modulations on the incoming photons coupled with physics-informed optimization to extract more information from the available signal than traditional methods. Experimentally, our approach achieves a target PSNR of 24 dB with just tens of noisy measurements, under severe turbulence up to D/r0~24. This robust performance in photon-starved conditions unlocks new capabilities in space situational awareness and remote sensing.