Scholarlink

Objective <p>Enhancing chest X-ray (CXR) images is crucial for accurate clinical diagnoses. Conventional methods, like histogram equalization, often wash out details and cause intensity shifts. This study proposes a novel exposure region-based modified adaptive histogram equalization (ERBMAHE) technique to address these limitations.</p> Methods <p>ERBMAHE divides CXR images into underexposed, well-exposed, and overexposed regions using our group’s 9IEC algorithm. The well-exposed region is further subdivided, creating five histograms. A novel adaptive probability density function (PDF) and power-law transformation are applied, enhancing contrast through modified HE equations. A fast local Laplacian filter improves detail while maintaining uniform illumination.</p> Results <p>Tested on 600 Kaggle CXR images, ERBMAHE outperformed six techniques, demonstrating improved illumination, detail enhancement, and naturalness. It achieved higher metrics such as discrete entropy (DE), peak signal to noise ratio (PSNR), absolute mean brightness error (AMBE), and structural similarity index (SSIM), with qualitative analysis validated by a doctor’s survey.</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

Contrast enhancement in chest X-ray imaging using exposer region-based modified adaptive histogram equalization (ERBMAHE)

  • 期刊论文

  • 发表日期：

    2025年7月10日

查看全文
  • Shivam Gangwar,
  • Reeta Devi,
  • Nor Ashidi Mat Isa

摘要

Objective

Enhancing chest X-ray (CXR) images is crucial for accurate clinical diagnoses. Conventional methods, like histogram equalization, often wash out details and cause intensity shifts. This study proposes a novel exposure region-based modified adaptive histogram equalization (ERBMAHE) technique to address these limitations.

Methods

ERBMAHE divides CXR images into underexposed, well-exposed, and overexposed regions using our group’s 9IEC algorithm. The well-exposed region is further subdivided, creating five histograms. A novel adaptive probability density function (PDF) and power-law transformation are applied, enhancing contrast through modified HE equations. A fast local Laplacian filter improves detail while maintaining uniform illumination.

Results

Tested on 600 Kaggle CXR images, ERBMAHE outperformed six techniques, demonstrating improved illumination, detail enhancement, and naturalness. It achieved higher metrics such as discrete entropy (DE), peak signal to noise ratio (PSNR), absolute mean brightness error (AMBE), and structural similarity index (SSIM), with qualitative analysis validated by a doctor’s survey.