Scholarlink

We discuss a learning-based low-rank (LLR) denoising method with an application to ultrasound images. Given a training data set of ground truth images (Fig. 4.1), we apply different types of noise and intensities (e.g., Gaussian, salt and pepper, Poisson, exponential, speckle) and extract 3D blocks through the block matching algorithm [5]. Then, LLR computes the optimal singular values through a proper optimisation applied to the SVD of each 3D block. LLR iterates this approach, where the input image of each iteration is the denoised image at the previous step.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Learning-Based Low-Rank Denoising

  • 章节

  • 发表日期：

    2026年

查看全文
  • Simone Cammarasana,
  • Giuseppe Patanè

摘要

We discuss a learning-based low-rank (LLR) denoising method with an application to ultrasound images. Given a training data set of ground truth images (Fig. 4.1), we apply different types of noise and intensities (e.g., Gaussian, salt and pepper, Poisson, exponential, speckle) and extract 3D blocks through the block matching algorithm [5]. Then, LLR computes the optimal singular values through a proper optimisation applied to the SVD of each 3D block. LLR iterates this approach, where the input image of each iteration is the denoised image at the previous step.