Correcting Optical Aberration via Depth-Aware Point Spread Functions.

Area of Science:

• Computer Vision
• Optical Engineering
• Image Processing

Background:

• Optical aberrations are common in lens-based imaging, causing blur and chromatic distortions.
• These aberrations vary with incident light angles and lens properties, complicating image quality.

Purpose of the Study:

• To develop a transferable image simulation system for simple lenses.
• To enable effective optical aberration correction using computational imaging principles.

Main Methods:

• Designed a simulation system using multi-wavelength, depth-aware, spatially-variant four-dimensional point spread functions (4D-PSFs).
• Developed a multi-scale optical aberration correction network with depth estimation and image restoration branches.
• Employed adaptive filters, dynamic convolutions, and self-attention mechanisms for spatially-variant restoration.

Main Results:

• The simulation system effectively models optical aberrations by adjusting lens-dependent parameters.
• The proposed network achieved favorable performance against state-of-the-art methods in degraded image reconstruction.
• Multi-scale feature extraction enhanced fine details and contextual information.

Conclusions:

• The developed system and network provide an effective approach for simulating and correcting optical aberrations.
• This method alleviates the need for extensive dataset collection in optical aberration correction.