Random resampling masks: a non-Bayesian one-shot strategy for noise reduction in digital holography

Area of Science:

• Optics and Photonics
• Digital Image Processing
• Computational Imaging

Background:

• Holographic imaging is susceptible to degradation from speckle and incoherent additive noise.
• Existing Bayesian denoising methods require prior knowledge of noise statistics.
• One-shot noise reduction is desirable for simplified and faster holographic recording, but often results in resolution loss.

Purpose of the Study:

• To propose a fast, non-Bayesian denoising method for holographic imaging.
• To achieve significant incoherent noise reduction without prior knowledge of noise statistics.
• To avoid the resolution loss typically associated with one-shot denoising techniques.

Main Methods:

• A novel non-Bayesian denoising approach using numerical synthesis of a moving diffuser.
• Generation of multiple uncorrelated reconstructions from a single hologram via random complementary resampling masks.
• Implementation of a fast computational method for noise reduction.

Main Results:

• Demonstrated significant reduction in incoherent noise, approaching theoretical improvement bounds.
• Achieved substantial improvement in holographic image contrast.
• Preserved the original resolution of the unprocessed holographic images.

Conclusions:

• The proposed method offers an effective solution for denoising holographic images in a single acquisition.
• It overcomes the resolution-degradation trade-off of previous one-shot methods.
• This technique enhances the practical applicability of holographic imaging by improving image quality and simplifying the recording process.