Imaging operator in indirect imaging correlography.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning

Background:

• Indirect imaging correlography (IIC) is a coherent imaging technique for non-line-of-sight (NLOS) imaging.
• IIC recovers sub-millimeter resolution images of obscured objects at large standoffs.
• Predicting IIC resolution in complex NLOS scenes is challenging due to object position and pose variations.

Purpose of the Study:

• To develop a mathematical model for the IIC imaging operator in NLOS scenarios.
• To derive and experimentally validate expressions for spatial resolution based on scene parameters.
• To propose a self-supervised deep neural network for reconstructing images from autocorrelation data.

Main Methods:

• Development of a mathematical model for the indirect imaging correlography (IIC) imaging operator.
• Derivation of spatial resolution expressions as a function of object position and pose.
• Experimental validation of the derived resolution formulas.
• Implementation of a self-supervised deep neural network for image reconstruction from autocorrelation.

Main Results:

• A validated mathematical model accurately predicts IIC imaging operator behavior in NLOS scenes.
• Spatial resolution is quantitatively linked to object position and pose parameters.
• Successful reconstruction of objects with ~250 μm features at 1m standoffs using the deep learning framework.

Conclusions:

• The developed mathematical model enhances the prediction of IIC performance in NLOS imaging.
• The study provides a method to optimize IIC resolution based on scene geometry.
• The proposed deep learning approach demonstrates effective image reconstruction for obscured objects in NLOS imaging.