Influence of diffraction distance on image restoration in deep learning networks

Area of Science:

• Computational Imaging
• Deep Learning
• Image Restoration

Background:

• Deep learning has advanced computational imaging.
• Few studies address diffraction distance's impact on deep learning-based image restoration.

Purpose of the Study:

• Investigate the effect of diffraction distance on image restoration using PhysenNet.
• Provide a theoretical framework for diffraction images and propagators.
• Determine optimal propagators for different diffraction distances.

Main Methods:

• Utilized the PhysenNet neural network for training on diffraction images.
• Experimented with various diffraction distances and propagators.
• Analyzed the impact of different propagators on network performance.

Main Results:

• Established a theoretical framework for diffraction imaging.
• Identified optimal propagators for image recovery at specific diffraction distances.
• Demonstrated the influence of diffraction distance on PhysenNet performance.

Conclusions:

• Diffraction distance significantly impacts deep learning-based image restoration.
• Optimal propagators can be determined for various diffraction distances.
• Findings expand the application of neural networks in computational imaging.