Scholarlink

<p>Outdoor images often suffer from diminished visibility caused by haze, a type of atmospheric attenuation that scatters light due to particles such as water droplets, dust, and smoke. This phenomenon leads to blurry and faded images, posing significant challenges in computer vision applications. Image dehazing aims to remove haze while preserving the fine details and edges of objects within the image. This paper proposes an Enhanced Lower Bound Transmission Dehazing (LBTD) technique that integrates advanced preprocessing methods to improve image quality before applying the dehazing process. These preprocessing steps include the homomorphic process, Contrast Limited Adaptive Histogram Equalization (CLAHE), and a fast dehazing method. These techniques minimize noise and enhance contrast in hazy images prior to employing the LBTD algorithm. The LBTD technique relies on a refined lower bound on transmission, determined through a bounding function (BF) and a quality control parameter. Using a nonlinear model, the BF is estimated for more accurate transmission prediction. By evaluating the differences between the minimum color channels of hazy and clear images, the proposed method reduces reconstruction errors during dehazing. The fast dehazing process incorporates multi-exposure image fusion, guided filtering, and gamma correction, which collectively enhance image quality by reducing brightness in hazy areas without compromising detail. The performance of the proposed technique is evaluated on videos captured in the visible and near-infrared (NIR) spectrums, demonstrating superior results compared to standalone LBTD and alternative dehazing methods. Quantitative assessment is performed using Peak Signal-to-Noise Ratio (PSNR), correlation metrics, entropy values, Feature Similarity Index Measurement (FSIM), and FSIM for chromatic components (FSIMC). Additionally, spectral entropy analysis and histogram evaluations confirm the efficacy of the proposed technique. The results indicate enhancement percentages of 55.19% for visible images and 38.07% for NIR images, based on PSNR values, highlighting the robustness and versatility of the approach.</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

Video and image quality enhancement using an enhanced lower bound on transmission map dehazing technique

  • 期刊论文

  • 发表日期：

    2025年3月15日

查看全文
  • Abeer Ayoub,
  • Walid El-Shafai,
  • Fathi E. Abd El-Samie,
  • Ehab K. I. Hamad,
  • S. EL-Rabaie

摘要

Outdoor images often suffer from diminished visibility caused by haze, a type of atmospheric attenuation that scatters light due to particles such as water droplets, dust, and smoke. This phenomenon leads to blurry and faded images, posing significant challenges in computer vision applications. Image dehazing aims to remove haze while preserving the fine details and edges of objects within the image. This paper proposes an Enhanced Lower Bound Transmission Dehazing (LBTD) technique that integrates advanced preprocessing methods to improve image quality before applying the dehazing process. These preprocessing steps include the homomorphic process, Contrast Limited Adaptive Histogram Equalization (CLAHE), and a fast dehazing method. These techniques minimize noise and enhance contrast in hazy images prior to employing the LBTD algorithm. The LBTD technique relies on a refined lower bound on transmission, determined through a bounding function (BF) and a quality control parameter. Using a nonlinear model, the BF is estimated for more accurate transmission prediction. By evaluating the differences between the minimum color channels of hazy and clear images, the proposed method reduces reconstruction errors during dehazing. The fast dehazing process incorporates multi-exposure image fusion, guided filtering, and gamma correction, which collectively enhance image quality by reducing brightness in hazy areas without compromising detail. The performance of the proposed technique is evaluated on videos captured in the visible and near-infrared (NIR) spectrums, demonstrating superior results compared to standalone LBTD and alternative dehazing methods. Quantitative assessment is performed using Peak Signal-to-Noise Ratio (PSNR), correlation metrics, entropy values, Feature Similarity Index Measurement (FSIM), and FSIM for chromatic components (FSIMC). Additionally, spectral entropy analysis and histogram evaluations confirm the efficacy of the proposed technique. The results indicate enhancement percentages of 55.19% for visible images and 38.07% for NIR images, based on PSNR values, highlighting the robustness and versatility of the approach.