A Denoising Method for Randomly Clustered Noise in ICCD Sensing Images Based on Hypergraph Cut and Down Sampling.

Area of Science:

• Image processing
• Low-light imaging technology
• Scientific instrumentation

Background:

• Intensified charge-coupled device (ICCD) sensors capture images in extremely low-light conditions.
• ICCD images often suffer from spatially clustered noise that general filtering methods struggle to address.
• Existing denoising techniques are often ineffective against the specific noise patterns in ICCD imagery.

Purpose of the Study:

• To develop an effective denoising algorithm for removing randomly clustered noise from ICCD images.
• To preserve structural and textural information in low-light ICCD images.
• To improve the overall quality of ICCD images for scientific applications.

Main Methods:

• A novel approach identifies clustered noise by down-sampling, up-sampling, and comparing images.
• The algorithm over-segments images into flat patches, classifying them as noisy or noise-free using hypergraph cuts.
• Noise-free patches are processed with Block-Matching and 3D filtering (BM3D); noisy patches are corrected by subtracting identified noise, followed by sparse noise reduction using robust principal component analysis.

Main Results:

• The proposed algorithm successfully removes randomly clustered noise from ICCD images.
• The method effectively preserves true textural and structural information within the images.
• Experimental comparisons demonstrate superior performance over four existing denoising algorithms.

Conclusions:

• The developed algorithm provides a robust solution for denoising ICCD images with clustered noise.
• This technique enhances the utility of ICCD sensors for scientific imaging in challenging low-light environments.
• The findings contribute to advancing image quality in scientific instrumentation and low-light photography.