Scholarlink

<p>The traditional Endoscopy method is a golden standard technique for observing gastrointestinal tract diseases. To properly identify gastrointestinal diseases, doctors mostly need good-quality images. The quality of the endoscopic images is mostly low in different hospitals due to the limited camera performance, complex gastrointestinal (GI) tract environment, and poor illumination. Most of the existing techniques often use the information of the image or multiple images of the same scene to offer enhancement. To overcome the issues associated with low contrast and noisy endoscopic images, this paper presents a novel image enhancement technique mainly developed to improve diagnostic accuracy and GI tract visualization. The proposed model begins by converting the images to the Hue Saturation Value (HSV) space to independently process the intensity component. In this proposed study, we applied a half-unit Weighted Bilinear Interpolation (Hu-WBI) algorithm method combined with the Contrast Limited Adaptive Histogram Equalization (CLAHE) on the inverted image intensity component by optimizing contrast while controlling noise. To achieve seamless contrast enhancement, the Hu-WBI is applied to overcome the over-enhanced images with artifact issues associated with CLAHE. To enhance the sharpness of the final images, an unsharp mask filter integrated with a Super Resolution Generative Adversarial Network (SRGAN) is applied. The hybrid methodoly preserves a details of high-frequency and it reduces excessive noise amplification Since the normalized intensity component is removed from the Low Pass filter, the final enhanced GI images are clearer and more precise which is widely adopted clinical technique for reading in the Computer Aided Diagnostic system. The proposed model is deployed in the Internet of Things (IoT) based cloud platform to improve the diagnostic capabilities of the medical images and demonstrate its practical impact on telemedicine. The proposed method achieves a very low error rate under extensive simulation experiments. The simulation results were conducted on two different datasets utilizing different performance evaluation metrics such as different metrics such as Structural Similarity Index (SSIM), Mean Squared Error (MSE), Adjusting image Intensity values (ADJ) and Feature Similarity Index (FSIM). When compared to the existing techniques, the proposed model improves the FSIM and IRMLE values by up to 32%. The proposed model offers a PSNR, SSIM, and LPIPS value of 20.1212, 0.8945, and 0.1456 when evaluated using the CVC clinic dataset.</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 unsharp masking filter technique with weighted bilinear interpolation for endoscopy image quality and sharpness enhancement

  • 期刊论文

  • 发表日期：

    2026年4月27日

查看全文
  • Manikandan Jagarajan,
  • Ramkumar Jayaraman

摘要

The traditional Endoscopy method is a golden standard technique for observing gastrointestinal tract diseases. To properly identify gastrointestinal diseases, doctors mostly need good-quality images. The quality of the endoscopic images is mostly low in different hospitals due to the limited camera performance, complex gastrointestinal (GI) tract environment, and poor illumination. Most of the existing techniques often use the information of the image or multiple images of the same scene to offer enhancement. To overcome the issues associated with low contrast and noisy endoscopic images, this paper presents a novel image enhancement technique mainly developed to improve diagnostic accuracy and GI tract visualization. The proposed model begins by converting the images to the Hue Saturation Value (HSV) space to independently process the intensity component. In this proposed study, we applied a half-unit Weighted Bilinear Interpolation (Hu-WBI) algorithm method combined with the Contrast Limited Adaptive Histogram Equalization (CLAHE) on the inverted image intensity component by optimizing contrast while controlling noise. To achieve seamless contrast enhancement, the Hu-WBI is applied to overcome the over-enhanced images with artifact issues associated with CLAHE. To enhance the sharpness of the final images, an unsharp mask filter integrated with a Super Resolution Generative Adversarial Network (SRGAN) is applied. The hybrid methodoly preserves a details of high-frequency and it reduces excessive noise amplification Since the normalized intensity component is removed from the Low Pass filter, the final enhanced GI images are clearer and more precise which is widely adopted clinical technique for reading in the Computer Aided Diagnostic system. The proposed model is deployed in the Internet of Things (IoT) based cloud platform to improve the diagnostic capabilities of the medical images and demonstrate its practical impact on telemedicine. The proposed method achieves a very low error rate under extensive simulation experiments. The simulation results were conducted on two different datasets utilizing different performance evaluation metrics such as different metrics such as Structural Similarity Index (SSIM), Mean Squared Error (MSE), Adjusting image Intensity values (ADJ) and Feature Similarity Index (FSIM). When compared to the existing techniques, the proposed model improves the FSIM and IRMLE values by up to 32%. The proposed model offers a PSNR, SSIM, and LPIPS value of 20.1212, 0.8945, and 0.1456 when evaluated using the CVC clinic dataset.