Noise-robust pixel-super-resolved multi-image phase retrieval with coherent illumination

Multi-image phase retrieval (MPR) algorithms play a significant role in the success of coherent diffractive imaging and in-line lensfree digital holographic imaging. However, they still suffer from the pixelation problem. By using them in combination with pixel super-resolution algorithms, the problem could be partially alleviated. But the existing pixel-super-resolved MPR method requires delicate parameter tuning and additional de-noising techniques. In this paper, we propose a new noise-robust pixel-super-resolved MPR scheme. It obviates the need for the mechanical movement of traditional pixel super-resolution methods and needs no extra measurements so that it can be directly applied in the existing system or dataset. Compared with the existing reconstruction method, it requires no sophisticated parameter selection and exhibits higher noise robustness, especially under laser illumination. Also, a new and effective focus measure is put forward so as to locate the object plane with higher precision and robustness. As a whole, our scheme provides a new label-free way of complex field imaging with a large field of view and high spatial resolution.