Reproduce “Recursive Threshold Median Filter and Autoencoder for Salt-and-Pepper Denoising: SSIM analysis of Images and Entropy Maps” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Petr Boriskov, Kirill Rudkovskii, Andrei Velichko
Researchers at Petrozavodsk State University introduced recursive threshold median filtering (2MF and MFs-AE) for salt-and-pepper denoising, achieving robust performance under strong noise conditions while maintaining computational efficiency. They also developed SSIM_Map, a novel image quality metric based on entropy maps, providing more sensitive evaluation of blur and structural integrity.
Key insights
- Recursive thresholding dramatically improves the Median Filter's performance for SP noise removal, reaching near-perfect restoration at low noise levels and marked improvement for strong corruption (>60% noise).
- A simple three-layer autoencoder is ineffective as a standalone denoiser for strong noise but proves useful for feature aggregation within multi-level schemes like MFs-AE.
- SSIM_Map is more sensitive to blur, local intensity transitions, and fine detail preservation than standard SSIM_Img, offering a valuable tool for objective assessment and parameter tuning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.