Padding-enabled real-time high-fidelity temporal single pixel imaging

Abstract One-time readout temporal ghost imaging (TGI) embodies a coveted single pixel imaging in time domain. However, false positives impact the retrieved images negatively and it is slow by design building on the cross-correlation of massive data. Our sparse recovery more than addresses these issues by eliminating one of two arms central to TGI, which allows moving beyond compressive TGI. Real-time noise-free high-fidelity temporal single pixel imaging is achieved by implementing a sacrificial time pad that flexibly and efficiently absorbs a sparsity-defying bias existent even under structured illumination. A factor of ≈ 3 × 10 3 improvement is obtained in terms of frame rate.