Scholarlink

<p>Image denoising remains a foundational task in high-precision imaging fields such as medical diagnostics and industrial defect analysis. Classical techniques often fall short in preserving structural and directional features, while deep learning models may lack generalizability across diverse imaging domains. This review presents a consolidated perspective on a two-phase denoising architecture that synergizes handcrafted transform-domain preprocessing with AI-based refinement. Phase I employs a Sliced Ridgelet Transform (SRT) pipeline incorporating Radon projections, adaptive thresholding, and bit-plane entropy control to yield sparsified and edge-aware image representations. Phase II builds upon this by integrating residual U-Nets, attention mechanisms, and GAN-based modules to reconstruct fine textures and perceptual details. The review synthesizes literature evolution, methodology, evaluation metrics, and experimental validations, establishing the proposed dual-phase framework as a robust and generalizable solution for real-world denoising applications.</p>

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

A Review on AI-augmented Dual-phase Denoising Using Sliced Ridgelet Transform and Deep Learning for Medical and Industrial Image Restoration

  • 期刊论文

  • 发表日期：

    2026年4月10日

查看全文
  • Brahmanandam Sharath G,
  • Krishnanaik Vankdoth

摘要

Image denoising remains a foundational task in high-precision imaging fields such as medical diagnostics and industrial defect analysis. Classical techniques often fall short in preserving structural and directional features, while deep learning models may lack generalizability across diverse imaging domains. This review presents a consolidated perspective on a two-phase denoising architecture that synergizes handcrafted transform-domain preprocessing with AI-based refinement. Phase I employs a Sliced Ridgelet Transform (SRT) pipeline incorporating Radon projections, adaptive thresholding, and bit-plane entropy control to yield sparsified and edge-aware image representations. Phase II builds upon this by integrating residual U-Nets, attention mechanisms, and GAN-based modules to reconstruct fine textures and perceptual details. The review synthesizes literature evolution, methodology, evaluation metrics, and experimental validations, establishing the proposed dual-phase framework as a robust and generalizable solution for real-world denoising applications.