Scholarlink

<p>Tuberculosis (TB) is a potentially fatal bacterial disease that mostly affects the lungs but has the potential to extend to other body parts. Noise in TB imaging databases can cause erroneous diagnosis results and misinterpretations. To overcome this a novel Sputum image denoise using multi-scale deep analysis model and vector medical filter (SD-MDAVM) has been proposed for denoising the image dataset. The input Sputum images are initially processed to Multi Scalar Image Generation and Noice pixel Detection. The Noise Pixel Detector identifies the noisy pixels in the image, whereas the Multi Scale Image Generation generates versions of the images at various scales. Then the noisy images from the multi-scale generation and the Noisy Pixel detection are given to a MultiScale Deep Analysis Model based Filter (MSDAMF) to reduce the image noise. Finally, the noise free images are gathered from the MSDAMF and the Vector Mediant Filter. For the KTI-DB database, the suggested approach achieves the lowest MSE values of 65.476, 83.192, and 78.358 for images KTI-DB-0, KTI-DB-1, and KTI-DB-2, respectively. The SD-MDAVM approach achieves the lowest time consumption across all databases with 3.52, 3.75, and 3.64 for ZNSM-DB, KTI-DB, and LCH-DB, respectively.</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

Enhancing sputum image clarity with multi-scale deep analysis and vector median filtering

  • 期刊论文

  • 发表日期：

    2025年3月11日

查看全文
  • A. Amala Shiny,
  • B. Sivagami

摘要

Tuberculosis (TB) is a potentially fatal bacterial disease that mostly affects the lungs but has the potential to extend to other body parts. Noise in TB imaging databases can cause erroneous diagnosis results and misinterpretations. To overcome this a novel Sputum image denoise using multi-scale deep analysis model and vector medical filter (SD-MDAVM) has been proposed for denoising the image dataset. The input Sputum images are initially processed to Multi Scalar Image Generation and Noice pixel Detection. The Noise Pixel Detector identifies the noisy pixels in the image, whereas the Multi Scale Image Generation generates versions of the images at various scales. Then the noisy images from the multi-scale generation and the Noisy Pixel detection are given to a MultiScale Deep Analysis Model based Filter (MSDAMF) to reduce the image noise. Finally, the noise free images are gathered from the MSDAMF and the Vector Mediant Filter. For the KTI-DB database, the suggested approach achieves the lowest MSE values of 65.476, 83.192, and 78.358 for images KTI-DB-0, KTI-DB-1, and KTI-DB-2, respectively. The SD-MDAVM approach achieves the lowest time consumption across all databases with 3.52, 3.75, and 3.64 for ZNSM-DB, KTI-DB, and LCH-DB, respectively.