Ptychographical iterative engine based on pixel binning for improving the computational efficiency

Abstract

Ptychographical iterative engine (PIE) is an attractive modality of phase retrieval that can provide the quantitative phase of the sample and extend the field of view. For a large sample, a great deal of images with a large dataset are recorded, which will result in high requirements for the computing power and increase the calculation burden. Here, we propose a pixel binning strategy to improve the computational efficiency and reduce the calculation time of PIE. In this method, the recorded image chosen as the amplitude constraint in reconstruction algorithm is compressed by merging the amplitude values of the adjacent positions into one value by the linear superposition, and the compressed pattern is set as a new amplitude constraint to reconstruct the compressed object and probe with fewer pixel number and larger grid size compared to the uncompressed results. Then, the compressed values are extended by the interpolation method. The improvement in the computational efficiency at the point lies in the fact that it takes less time to do the free-space diffraction propagation calculation for images with fewer pixel number. Experiments demonstrate that the proposed method behaves good performance with high computational efficiency. The proposed approach would be helpful for large-scale imaging with high computational efficiency.