Momentum-Accelerated Incoherent Fourier Ptychographic Photography

Incoherent Fourier Ptychography (IFP) is a super-resolution imaging method, which requires lots of iteration time to recover the high-resolution object image from raw images. To improve the convergence rate of the IFP, we propose an improved algorithm, termed momentum-accelerated Incoherent Fourier Ptychography (mIFP). Based on the IFP algorithm, the mIFP adds momentum that is commonly used to accelerate the training of weights in a neural network. During the update process, the current vector is updated based on the downward trend of the objective function during several updates and the previous vector. The updated vector is then added to the objective function. We demonstrate the proposed approach by both simulations and experiments. By reducing the recovery time, mIFP may accelerate the applications of IFP.