融合零参考深度曲线的低照度图像增强与去噪算法
DOI:
作者:
作者单位:

中国人民公安大学信息网络安全学院

作者简介:

通讯作者:

中图分类号:

TP391.7

基金项目:

中国人民公安大学安全防范工程双一流专项(2023SYL08)


A low-light image enhancement and denoising algorithm incorporating zero reference depth curves
Author:
Affiliation:

1.School of Information Network Security, People'2.'3.s Public Security University of China

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为解决低光照条件下的图像噪声多、亮度低、细节模糊等问题,提出了一种融合零参考深度曲线的低光照图像增强与去噪算法(UMDCEAD-NET)。该算法首先在u-net的下采样模块中融入mobile-net模块,以实现有效的特征提取,从而保留更多的图像细节信息。其次,利用深度曲线估计(LE-曲线),解决图像像素级光照不足的问题。最后,结合AD-NET算法对增强后的图像进行降噪处理,以使得图像更符合人眼的视觉感知。实验结果显示,所提出的算法在公开数据集Zero-DCE平均峰值信噪比(PSNR)达到22.29,相比Zero-DCE++算法提高了32%,在公开数据集LOL平均峰值信噪比(PSNR)达到21.15,相比SGZ算法提高了3%。可见,该算法能够有效地解决增强后图像的噪声问题,使得增强后的图像暗部和亮部区域的细节信息更加丰富,与其他主流算法相比该算法图像质量有明显提升。

    Abstract:

    In order to solve the problems of high image noise, low brightness and blurred details under low-light conditions, a low-light image enhancement and denoising algorithm (UMDCEAD-NET) incorporating a zero reference depth curve is proposed. The algorithm firstly incorporates the mobile-net module into the downsampling module of u-net to achieve effective feature extraction so as to retain more image detail information. Secondly, depth curve estimation (LE-curve) is utilised to solve the problem of insufficient illumination at the pixel level of the image. Finally, the enhanced image is combined with the AD-NET algorithm for noise reduction in order to make the image more compatible with the visual perception of the human eye. The experimental results show that the proposed algorithm achieves an average peak signal-to-noise ratio (PSNR) of 22.29 in the public dataset Zero-DCE, which is 32% higher than that of the Zero-DCE++ algorithm, and an average peak signal-to-noise ratio (PSNR) of 21.15 in the public dataset LOL, which is 3% higher than that of the SGZ algorithm. It can be seen that the algorithm can effectively solve the noise problem of the enhanced image, so that the enhanced image dark and bright regions of the detailed information is richer, compared with other mainstream algorithms the algorithm image quality has been significantly improved.

    参考文献
    相似文献
    引证文献
引用本文

田博文,丁建伟,户子睿. 融合零参考深度曲线的低照度图像增强与去噪算法[J]. 科学技术与工程, , ():

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2023-11-14
  • 最后修改日期:2024-05-25
  • 录用日期:2024-06-05
  • 在线发布日期:
  • 出版日期:
×
亟待确认版面费归属稿件,敬请作者关注