Scholarlink

<p>High-quality medical imaging is fundamental to accurate diagnosis and effective clinical decision-making, yet real-world acquisitions are frequently degraded by noise, artifacts, and loss of detail. To address these challenges, we propose a novel non-convex total variation (TV) regularization model that leverages a hyper-Laplacian prior to enhance sparsity and preserve structural edges. The model is formulated using an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\ell _p\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>ℓ</mi> <mi>p</mi> </msub> </math></EquationSource> </InlineEquation>-quasi norm (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(0&lt; p &lt; 1\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>0</mn> <mo>&lt;</mo> <mi>p</mi> <mo>&lt;</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation>) and is efficiently solved via the Alternating direction method of multipliers framework integrated with the Chebyshev iterative method. This hybrid optimization strategy decomposes the complex problem into tractable subproblems, significantly improving convergence speed and computational efficiency. Extensive experiments on real medical images with varying noise levels demonstrate that the proposed method consistently outperforms state-of-the-art techniques in PSNR, SSIM, edge preservation, and spectral fidelity. The results highlight the method’s robustness, precision, and practical value for clinical image enhancement.</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

Medical Image Denoising Using Non-Convex TV Regularization with Chebyshev-Optimized ADMM

  • 期刊论文

  • 发表日期：

    2025年11月29日

查看全文
  • Narendra Kumar,
  • Gaurav Bhatnagar

摘要

High-quality medical imaging is fundamental to accurate diagnosis and effective clinical decision-making, yet real-world acquisitions are frequently degraded by noise, artifacts, and loss of detail. To address these challenges, we propose a novel non-convex total variation (TV) regularization model that leverages a hyper-Laplacian prior to enhance sparsity and preserve structural edges. The model is formulated using an \(\ell _p\) ℓ p -quasi norm ( \(0< p < 1\) 0 < p < 1 ) and is efficiently solved via the Alternating direction method of multipliers framework integrated with the Chebyshev iterative method. This hybrid optimization strategy decomposes the complex problem into tractable subproblems, significantly improving convergence speed and computational efficiency. Extensive experiments on real medical images with varying noise levels demonstrate that the proposed method consistently outperforms state-of-the-art techniques in PSNR, SSIM, edge preservation, and spectral fidelity. The results highlight the method’s robustness, precision, and practical value for clinical image enhancement.