Adaptive sparse reconstruction for lensless digital holography via PSF estimation and phase retrieval

Area of Science:

• Optics and Photonics
• Digital Imaging
• Computational Imaging

Background:

• Lensless digital holography offers potential for various applications but faces challenges in reconstructing high-quality images due to phase information loss.
• Conventional reconstruction methods rely on inverse problems and are sensitive to mismatches between the imaging model and the actual system.

Purpose of the Study:

• To enhance the robustness of holographic image reconstruction algorithms.
• To introduce a novel method that learns system parameters directly from data.

Main Methods:

• Developed an adaptive sparse reconstruction (ASR) method for in-line lensless digital holography.
• ASR jointly performs holographic reconstruction, point spread function (PSF) estimation, and phase retrieval in an unsupervised manner.
• The method maximizes the sparsity of reconstructed images, using a sparsity prior and the image formation model.

Main Results:

• ASR demonstrates improved robustness compared to traditional reconstruction methods.
• Experimental results with synthetic and real data validate the effectiveness of ASR.
• The method excels in scenarios where the theoretical PSF deviates from the actual system's PSF.

Conclusions:

• Adaptive sparse reconstruction (ASR) offers a more robust approach to lensless digital holography.
• The unsupervised, data-driven PSF learning overcomes limitations of physics-based approximations.
• ASR provides a significant advantage for reconstructing high-quality holographic images, particularly in non-ideal imaging conditions.