Super-Resolution Parameter Estimation Using Machine Learning-Assisted Spatial Mode Demultiplexing.

Area of Science:

• Optics and Photonics
• Machine Learning Applications
• Computational Imaging

Background:

• Resolving fine details in optical imaging is often limited by the diffraction limit.
• Spatial mode demultiplexing (SPADE) offers a potential pathway to overcome these limitations.
• Machine learning (ML) presents opportunities for advanced data analysis in complex imaging scenarios.

Purpose of the Study:

• To develop and evaluate a machine learning model for estimating the separation and relative brightness of closely spaced light sources.
• To assess the model's performance in achieving sub-diffraction limit resolution.
• To explore the application of ML-assisted SPADE imaging for overcoming diffraction limitations.

Main Methods:

• Utilized a multi-planar light converter (MPLC) to perform SPADE imaging.
• Trained, validated, and tested a lightweight machine learning model on experimental laboratory data.
• Focused on estimating source separation and relative brightness below the diffraction limit.

Main Results:

• The ML model accurately estimated source separation up to two orders of magnitude below the diffraction limit for comparable brightness sources.
• Achieved accurate sub-diffraction separation resolution even when source brightness differed by four orders of magnitude.
• Performance was limited by cross-talk within the MPLC.

Conclusions:

• ML-assisted SPADE imaging demonstrates significant potential for achieving sub-diffraction resolution.
• The developed ML model shows promise for applications in astronomical imaging.
• Further improvements in MPLC technology could enhance the capabilities of this technique.