Blind coherent modulation imaging using momentum acceleration and sample priors

Abstract

Coherent Modulation Imaging (CMI) stands out as a novel lensless imaging technique with notable advantages such as rapid convergence and single-shot capability. Nevertheless, conventional CMI implementations necessitate an additional step to acquire prior information about the modulator function, introducing complexity and reliance on other imaging techniques. Previous attempts to mitigate the requirement for precise modulator information using diverse objects have encountered slow convergence speeds. Here, we present an improved CMI algorithm, termed as blind CMI, which achieves blind recovery without prior knowledge of the modulator. This is achieved by leveraging sample priors and incorporating momentum acceleration. We validate our method through numerical simulations and optical experiments, demonstrating that the proposed blind CMI outperforms other state-of-the-art methods in terms of both convergence speed and reconstruction quality.