Model-based autofocus for near-field phase retrieval.

Area of Science:

• Coherent lensless microscopic imaging
• Phase retrieval algorithms
• Computational imaging

Background:

• The phase problem in lensless microscopy leads to ill-posed reconstructions.
• Accurate estimation of the Fresnel number is crucial for autofocusing and sharp image reconstruction.
• Existing autofocusing methods are sensitive to noise and neglect algorithmic properties.

Purpose of the Study:

• To develop a novel autofocusing criterion for lensless microscopy.
• To improve the accuracy of Fresnel number estimation.
• To automate the autofocusing process for in-situ/operando experiments.

Main Methods:

• A novel model-matching criterion is proposed.
• An autofocusing framework is derived using a phase-retrieval approach.
• A downhill-simplex method is employed for automatic Fresnel number optimization.

Main Results:

• The proposed criterion improves autofocusing by considering the reconstruction algorithm, forward model, and hologram data.
• The framework demonstrates robustness across different datasets.
• Accurate Fresnel number estimation leads to sharper, artifact-free images.

Conclusions:

• The novel model-matching approach offers a robust solution for autofocusing in lensless microscopy.
• This method enhances image reconstruction quality by accurately estimating the Fresnel number.
• The automated framework is suitable for demanding experimental conditions.