High-resolution optical microscopy in complex environments with a single-pixel detector

Optical microscopy faces a challenge in strongly scattering environments due to severe light attenuation and wave degradation. Here, we report high-resolution optical microscopy in complex environments with a single-pixel detector. By projecting miniaturized random patterns onto a specimen, a series of light intensities can be synchronously collected via single-pixel detection. Dynamic variations in the turbidity in complex scattering environments induce nonlinear attenuations. A framework of untrained neural networks enhanced by a physical model is developed to estimate a series of scattering-induced scaling factors and achieve high-resolution object reconstruction. The designed optical microscopy system, employing a tunable lens with autofocusing, is also applied to reconstruct high-quality and high-resolution images of biological specimens over varying fields of view against complex and dynamic scattering. It is demonstrated in experiments that the proposed method is effective and robust, providing a viable approach for optical microscopy through complex scattering in dynamic media.