Scholarlink

<p>Multi-scale retinal enhancement and weighted homomorphic filtering algorithms are proposed to address the problems of weak texture features, unclear detail information, and uneven and low contrast of feature regions in Si<sub>3</sub>N<sub>4</sub> bearing roller images. Combined with the weak texture characteristics of Si<sub>3</sub>N<sub>4</sub>-bearing roller microcracks, this creates the multi-channel convolution equation for retinal color recovery. Break down the characteristics of an image into information across various scales and boost the contrast of the texture's less prominent features. Based on the different gray value components of different frequencies of the feature image, distinct gray value components exist according to the feature image's various frequencies. Set up the Gaussian difference equation, extend local grayscale values, and realize noise removal of texture features`. The average PSNR of the optimized microcrack weak texture feature image is 21.4556&#xa0;dB, as per the results. By comparing the experiments, the image entropy is improved by 23.8% on average, which effectively enhances the roller microcrack contrast and details of the texture feature image, improves the accuracy and quality of feature images.</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

Multi-scale retina enhancement paired with weighted homomorphic filtering as a combined picture improvement technique for Si3N4 bearing roller microcrack weak texture feature

  • 期刊论文

  • 发表日期：

    2025年5月13日

查看全文
  • Tao Chen,
  • Xin Xia,
  • Hui Yang,
  • Jianbo Le,
  • Weiwen Hu,
  • Nanxing Wu

摘要

Multi-scale retinal enhancement and weighted homomorphic filtering algorithms are proposed to address the problems of weak texture features, unclear detail information, and uneven and low contrast of feature regions in Si3N4 bearing roller images. Combined with the weak texture characteristics of Si3N4-bearing roller microcracks, this creates the multi-channel convolution equation for retinal color recovery. Break down the characteristics of an image into information across various scales and boost the contrast of the texture's less prominent features. Based on the different gray value components of different frequencies of the feature image, distinct gray value components exist according to the feature image's various frequencies. Set up the Gaussian difference equation, extend local grayscale values, and realize noise removal of texture features`. The average PSNR of the optimized microcrack weak texture feature image is 21.4556 dB, as per the results. By comparing the experiments, the image entropy is improved by 23.8% on average, which effectively enhances the roller microcrack contrast and details of the texture feature image, improves the accuracy and quality of feature images.