OCT Speckle Denoising with Diffusion Models

DDPM-based denoising for OCT medical images

Research assistant work at The Chinese University of Hong Kong, advised by Prof. Wu Yuan.

Overview

Multiplicative Speckle Noise Denoising of OCT Medical Images Based on Diffusion Models

This work develops a DDPM-based denoising method for speckle noise in optical coherence tomography (OCT) medical images to improve diagnostic accuracy.

Key Components

  • DDPM denoising: noise modeling and diffusion training pipelines for OCT speckle removal
  • Efficiency: enhanced convergence speed and prediction efficiency
  • Comprehensive benchmarks: reproduced DnCNN, U-Net, BM3D, TCFL-OCT, and DDPM baselines for comparative evaluation
  • Metrics: Python scripts computing PSNR, SSIM, and GCMSE
Denoising comparison. To be replaced with a figure from the project.

References