Scholarlink

<p>Optical coherence tomography (OCT), which is highly regarded, serves as a valuable instrument in both biological research and clinical diagnosis, and it is capable of providing high-resolution tissue images. However, the presence of speckle noise in OCT images can significantly reduce their quality, potentially leading to inaccurate study results and diagnostic procedures. To solve this problem, a fusion denoising algorithm based on ResNet and truncated Huber filtering is proposed. The source image is initially decomposed into base and detail layers using a truncated Huber filter. In order to achieve detail layer fusion, the multilayer features of the detail layer are initially extracted using a ResNet network. Subsequently, a set of candidate features is generated by applying the <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11227_2025_7690_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="13" /> </InlineMediaObject> <EquationSource Format="TEX">\({l}_{1}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>l</mi> <mn>1</mn> </msub> </math></EquationSource> </InlineEquation>-norm and weighted average rule. Then, a feature fusion scheme based on the average pixel intensity energy operator is employed to fuse these candidate features in order to achieve more accurate and comprehensive detail layer information fusion. For the base layer fusion, a fusion scheme based on spatial frequency similarity and structural similarity is proposed. Finally, the fused detail and base layers are synthesized to reconstruct the fused image. The results of experimental studies demonstrate that the algorithm exhibits comparable performance to some of the most advanced speckle noise removal techniques, as evidenced by both objective metrics and subjective visual inspection.</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

Image fusion and denoising of optical coherence tomography based on ResNet and truncated Huber filter

  • 期刊论文

  • 发表日期：

    2025年8月5日

查看全文
  • Huaiguang Chen,
  • Wenyu Wei,
  • Hehe Li,
  • Jing Gao

摘要

Optical coherence tomography (OCT), which is highly regarded, serves as a valuable instrument in both biological research and clinical diagnosis, and it is capable of providing high-resolution tissue images. However, the presence of speckle noise in OCT images can significantly reduce their quality, potentially leading to inaccurate study results and diagnostic procedures. To solve this problem, a fusion denoising algorithm based on ResNet and truncated Huber filtering is proposed. The source image is initially decomposed into base and detail layers using a truncated Huber filter. In order to achieve detail layer fusion, the multilayer features of the detail layer are initially extracted using a ResNet network. Subsequently, a set of candidate features is generated by applying the \({l}_{1}\) l 1 -norm and weighted average rule. Then, a feature fusion scheme based on the average pixel intensity energy operator is employed to fuse these candidate features in order to achieve more accurate and comprehensive detail layer information fusion. For the base layer fusion, a fusion scheme based on spatial frequency similarity and structural similarity is proposed. Finally, the fused detail and base layers are synthesized to reconstruct the fused image. The results of experimental studies demonstrate that the algorithm exhibits comparable performance to some of the most advanced speckle noise removal techniques, as evidenced by both objective metrics and subjective visual inspection.