Quantitative Phase Retrieval Through Scattering Medium via Compressive Sensing

Tong Peng,Xianghua Yu,Meiling Zhou,Chen Bai,Junwei Min,Baoli Yao,Runze Li,Chunmin Zhang,Dan Dan

doi:10.1109/jphot.2021.3136509

Abstract

Scattering media, such as biological tissues and turbid liquids, scatter light randomly and introduce several challenges when imaging objects behind them. The transmission matrix (TM) describes the relation between the input and output of a beam transmitted through a medium, which can be used to reconstruct a target located behind a scattering medium. However, the current TM methods cannot easily retrieve the phase distribution of objects inside or behind a scattering medium. In this work, a compressive sensing (CS) method to identify the TM of a scatter contained in an imaging system was investigated. By calibrating the TM, the phase information of the object can be retrieved quantitatively. This method allows one to retrieve multilevel and dynamic phase objects behind different scatters. The influence of the calibration parameters on the reconstruction quality was investigated in detail. The proposed method, featuring noninterference measurements of the TM and exploiting a large field of view, can be used in phase imaging applications.

Highlights

Research on focusing and imaging through a scattering medium has recently triggered substantial interest [1,2,3] due to its important applications in biomedical imaging, fire rescue, micromanipulation, etc. [4,5,6,7]
[31], which offers satisfactory accuracy while running hundreds of times faster with potentially parallel processes and acceleration by GPUs. phase retrieval vector approximate message passing (prVAMP) is a special case of the recently developed generalized vector approximate message passing (GVAMP) algorithm [33]. This is a typical algorithm for computing approximate minimum means squared error (MMSE) or maximum a posteriori (MAP) solution to inverse problems involving generalized linear measurements (GLMs); defined to be any measurement of the form y = F(z + w) with z = Φx where Φ is our measurement matrix, x is our signal of interest, w is noise, and F(·) denotes a simple nonlinear procedure, known as the generalized linear model, which evolved from compressive sensing (CS) [33]
The transmission matrix (TM) based on a generalized linear model of an imaging system with a scattering medium can be obtained by using a series of calibration images

Summary

INTRODUCTION

Research on focusing and imaging through a scattering medium has recently triggered substantial interest [1,2,3] due to its important applications in biomedical imaging, fire rescue, micromanipulation, etc. [4,5,6,7]. With the development of phase retrieval algorithms, the TM of a system with its scattering medium can be obtained by loading a series of calibrated phase images into a modulator and capturing the corresponding speckle patterns. By combining a digital micromirror device (DMD) and double phase retrieval (DPR), the TM of an imaging system can be learned effectively This method, featuring robust and avoiding interferometric measurements, can realize refocusing of light through scattering media [26,27]. Scattering medium can be effectively reconstructed, thereby performing phase retrieval with this approach does not provide extending the potential applications in the field of TM-based a quantitative result For this reason, only phase maps with imaging.

Principle of the method

Experimental setup

Inverse scattering with coherent illumination

Inverse scattering with partial coherent illumination

CONCLUSION