Scholarlink

<p>Most conventional mask-based edge detectors not only do produce thick and unclear edges in grayscale images, but they leave out some edges as well. In this paper, we develop a novel method that produces thin and clear edges in images. We analyze the effect of considering the regional information around a given pixel in edge detection. We treat each image as a set of 3<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\times\)</EquationSource> </InlineEquation>3 dark (low intensities) or bright (high intensities) regions depending on a prefixed threshold. Then, we estimate a set of pixels by linearly increasing or reducing the range of intensities of these regions following the kind of each one of them. Finally, the edge and direction maps are generated by a simple convolution process. We elaborate several edge detectors and experimentally demonstrate how each one of them can be applied according to requirements. Our proposal is tested for different types of images including standard and medical images such as CT and MRI and has proved to be competitive with the most known methods like Sobel and Canny.</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

An estimation method of regional information to improve mask-based edge detectors

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • A. Alifdal,
  • B. Bouda

摘要

Most conventional mask-based edge detectors not only do produce thick and unclear edges in grayscale images, but they leave out some edges as well. In this paper, we develop a novel method that produces thin and clear edges in images. We analyze the effect of considering the regional information around a given pixel in edge detection. We treat each image as a set of 3 \(\times\) 3 dark (low intensities) or bright (high intensities) regions depending on a prefixed threshold. Then, we estimate a set of pixels by linearly increasing or reducing the range of intensities of these regions following the kind of each one of them. Finally, the edge and direction maps are generated by a simple convolution process. We elaborate several edge detectors and experimentally demonstrate how each one of them can be applied according to requirements. Our proposal is tested for different types of images including standard and medical images such as CT and MRI and has proved to be competitive with the most known methods like Sobel and Canny.