Scholarlink

<p>Scattering artifacts in X-ray radiography can potentially reduce the quality of images obtained. This also has the effect of making extraction of useful image information in medical diagnostic and industrial non-destructive testing (NDT) challenging. This study proposes a combined approach by employing wavelet transform and convolutional neural networks (CNNs) to reduce scattering artifacts in X-ray radiography. Wavelet transform is used to decompose images into low and high frequency components to obtain their approximate and detailed features. Wavelet components are used as supervisory data to train CNN, which learns to extract the scattering affected input images and transform them into corrected output images. The proposed model shows acceptable performance in compensating X-ray scattering effects without losing essential structural details in radiographs. Training results confirm the methodology ability of optimizing image quality that is suitable for diverse medical or industrial X-ray imaging.</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

Hybrid deep learning and wavelet decomposition approach for X-ray scatter artifact correction

  • 期刊论文

  • 发表日期：

    2025年9月12日

查看全文
  • A. Taheri,
  • M. Askari

摘要

Scattering artifacts in X-ray radiography can potentially reduce the quality of images obtained. This also has the effect of making extraction of useful image information in medical diagnostic and industrial non-destructive testing (NDT) challenging. This study proposes a combined approach by employing wavelet transform and convolutional neural networks (CNNs) to reduce scattering artifacts in X-ray radiography. Wavelet transform is used to decompose images into low and high frequency components to obtain their approximate and detailed features. Wavelet components are used as supervisory data to train CNN, which learns to extract the scattering affected input images and transform them into corrected output images. The proposed model shows acceptable performance in compensating X-ray scattering effects without losing essential structural details in radiographs. Training results confirm the methodology ability of optimizing image quality that is suitable for diverse medical or industrial X-ray imaging.