GEOMScope: Large Field-of-view 3D Lensless Microscopy with Low Computational Complexity

Area of Science:

• Biomedical imaging
• Computational optics
• Microscopy

Background:

• Miniaturized, high-resolution 3D imaging is crucial for biomedical applications.
• Existing 3D imaging systems often use bulky lenses and mechanical refocusing, limiting their practicality.

Purpose of the Study:

• To introduce GEOMScope, a lensless, single-shot 3D microscope.
• To demonstrate its capability for high-resolution, real-time 3D reconstruction with reduced computational load.

Main Methods:

• Utilizes a single layer of a thin microlens array for image formation.
• Employs a novel reconstruction algorithm combining geometrical-optics-based pixel back projection and background suppression.
• Integrates geometrical optics principles with machine learning for 3D reconstruction.

Main Results:

• GEOMScope achieves high-resolution 3D imaging of resolution targets, fluorescent particles, and volumetric objects.
• Significantly reduces computational resources and increases reconstruction speed compared to other lensless imaging devices.
• Enables near real-time processing for imaging and recovering large-volume 3D objects.

Conclusions:

• GEOMScope offers a practical solution for miniaturized, high-speed, high-resolution 3D biomedical imaging.
• The joint design of optics and algorithms, integrating geometrical optics and machine learning, is key to its performance.
• Geometrical optics holds significant potential for advancing computational imaging technologies.