Signal-processing approaches for image-resolution restoration for TOMBO imagery

Area of Science:

• Optics and Photonics
• Computational Imaging
• Signal Processing

Background:

• Conventional optical systems use large apertures, limiting compactness.
• The Thin Observation Module by Bounded Optics (TOMBO) utilizes a lenslet array for compact, thin optical systems.
• TOMBO collects diverse low-resolution measurements, necessitating advanced methods for high-resolution image reconstruction.

Purpose of the Study:

• To develop and evaluate a computational method for resolution restoration using TOMBO measurements.
• To address the challenge of efficiently combining diverse low-resolution data into a high-resolution image.
• To assess the performance of the proposed method through simulations under various noise conditions.

Main Methods:

• Developed a computational data model based on Fourier optics.
• Proposed restoration algorithms minimizing Csiszár's I divergence, incorporating Poisson and Gaussian noise models.
• Adapted the expectation-maximization method for optimization, preserving nonnegativity constraints.
• Incorporated total variation regularization to mitigate artifacts.

Main Results:

• Simulations demonstrated the algorithm's ability to produce very high-quality estimates from noiseless data.
• Reasonably good estimates were achieved from noisy and incomplete measurements.
• The multiplicative expectation-maximization approach effectively handled nonnegativity constraints.

Conclusions:

• The developed computational method shows significant promise for high-resolution image reconstruction from TOMBO data.
• The approach is robust to noise and data incompleteness.
• Future work could explore measurement selection strategies for further optimization.