Embedded Processing for Extended Depth of Field Imaging Systems: From Infinite Impulse Response Wiener Filter to...

Area of Science:

• Optics and Photonics
• Digital Image Processing
• Computational Imaging

Background:

• Current methods for extending camera depth of field (DoF) often involve complex joint optimization of optical elements and digital processing with infinite deconvolution or neural networks.
• These techniques aim to improve imaging at greater distances or relax manufacturing tolerances for sensor positioning.

Purpose of the Study:

• To investigate depth of field (DoF) extension using embedded digital processing with a single finite convolution.
• To compare different finite impulse response (FIR) filter approaches, including learned and Wiener filter paradigms.

Main Methods:

• Developed an optical model for codesigned systems focused on DoF extension.
• Employed a Wiener filter paradigm to compute FIR filter coefficients, incorporating scene power spectral density (either learned or modeled).
• Compared various FIR filter designs and proposed a method for pre-optimizing filter sizes.

Main Results:

• Demonstrated the feasibility of DoF extension with single finite convolutions.
• Showcased that learned FIR filters offer adaptability to datasets, while Wiener-based filters provide comparable robustness.
• Presented a method for dimensioning FIR filter sizes before joint optimization.

Conclusions:

• Embedded processing with FIR filters is an effective strategy for depth of field extension.
• Both learned and Wiener filter approaches are viable, offering different advantages in terms of adaptability and robustness.
• The proposed dimensioning method aids in optimizing FIR filter performance for DoF extension systems.