Learning-based single-shot superresolution in diffractive imaging.

Area of Science:

• Optics and Photonics
• Machine Learning Applications
• Image Reconstruction

Background:

• Diffraction-limited imaging systems capture intensity images, which inherently limit resolution.
• Retrieving a superresolved object field from a single intensity image is a significant challenge in optical imaging.
• Existing methods often require multiple measurements or complex optical setups.

Purpose of the Study:

• To present a novel machine learning-based method for super-resolution object field retrieval.
• To overcome the resolution limitations imposed by diffraction in imaging systems.
• To demonstrate the method's effectiveness in lensless imaging scenarios, including those with scattering media.

Main Methods:

• A machine learning approach is employed to regress the inverse problem of diffractive imaging.
• The model is trained using pairs of object fields and their corresponding captured intensity images.
• The method utilizes a single captured intensity image for object field reconstruction.

Main Results:

• The proposed method successfully retrieves a superresolved object field from a single intensity image.
• Experimental validation demonstrates the effectiveness of the technique in a lensless imaging setup.
• The method shows robustness even in the presence of scattering media, improving image quality.

Conclusions:

• Machine learning offers a powerful tool for enhancing resolution in diffractive imaging.
• The developed method provides a pathway to achieve super-resolution from single intensity measurements.
• This technique has potential applications in various fields requiring high-resolution imaging, particularly in challenging environments.