High-fidelity lensless polarization imaging enabled by differentiable mask optimization and second-order regularized...

Area of Science:

• Optics and Photonics
• Computational Imaging
• Image Reconstruction

Background:

• Lensless coded aperture polarization imaging is promising for compact systems.
• Traditional designs suffer from modulation transfer function (MTF) zeros and noise amplification during reconstruction.
• Ill-posed reconstruction amplifies noise in nonlinear polarization computations.

Purpose of the Study:

• To develop a differentiable physical model for optimizing sub-aperture arrangement in coded apertures.
• To mitigate spectral zeros in the MTF and reduce noise in reconstructed images.
• To propose an advanced algorithm for solving the ill-posed reconstruction problem.

Main Methods:

• A differentiable physical model was developed to optimize sub-aperture spatial arrangement, avoiding contiguous transparent regions.
• This optimization eliminates MTF zeros and enhances mid-to-high frequency band energy.
• A total generalized variation momentum primal-dual hybrid gradient (TGV-MPDHG) algorithm was proposed, incorporating second-order regularization.

Main Results:

• The integrated methodology significantly improved reconstruction fidelity.
• Background noise was substantially reduced in the reconstructed images.
• Faster convergence and stable reconstruction quality were achieved across various object sizes.

Conclusions:

• The proposed differentiable physical model and TGV-MPDHG algorithm effectively address limitations in lensless coded aperture polarization imaging.
• This approach enhances image quality, reduces noise, and improves reconstruction stability.
• The findings pave the way for more robust and high-fidelity compact polarization imaging systems.