Deep learning colorful ptychographic iterative engine lens-less diffraction microscopy

Abstract

The ptychographic iterative engine (PIE) is an algorithm for coherent lens-less diffraction imaging. It facilitates the development of low-cost large field of view (FOV) lens-less microscopy, as it can expand the FOV extensively by moving the light source or the bio-medical pathological sample slide. In a PIE setup, the illumination light needs to have high/partially coherence. Thus, to get a colorful microscopy image, more than 3 illumination light sources with different dominant wavelengths are required in a traditional PIE setup. In this manuscript, an improved PIE based on computational deep learning color-transferring method is proposed to achieve colorful large FOV lens-less microscopy imaging. In our method, only one high/partially coherent light source is used, where the image data are three times less than those images under multiple illuminations for colorful PIE microscopy. It is believable that our colorful PIE microscopy enhanced by the deep color-transferring method would be helpful for the development of low-cost large FOV lens-less microscopes.