计算机与现代化

• 算法设计与分析 • 上一篇    下一篇

一种改进的雾天图像复原算法

  

  1. 长沙理工大学电气与信息工程学院,湖南长沙410114
  • 收稿日期:2015-01-30 出版日期:2015-04-27 发布日期:2015-04-29
  • 作者简介: 郭庚山(1987-),男,湖南郴州人,长沙理工大学电气与信息工程学院硕士研究生,研究方向:智能交通检测与控制; 叶青(1963-),女,湖南长沙人,教授,研究方向:智能交通 检测与控制; 胡鑫(1990-),男,陕西咸阳人,硕士研究生,研究方向:智能检测与模式识别。

 An Improved Haze Image Restoration Algorithm

  1. College of Electrical and Information Engineering, Changsha University of Science and Technology, Changsha 410114, China
  • Received:2015-01-30 Online:2015-04-27 Published:2015-04-29

摘要:

 针对传统暗通道先验算法对大气光亮度值A估算不够准确,图像复原后容易出现天空区域色彩失真这些问题,本文提出一种改进算法。通过设定3个天空阈值筛选出天空区域的像素,然后求
其亮度的均值得到大气光亮度值A,剔除白色物体的影响; 利用k-means聚类算法对分割天空区域和非天空区域的阈值实现自适应。该算法可以求得更加准确的大气光亮度值A,通过对分割阈值T自适应化
,可以更加精确地分割出天空区域,从而可以对天空区域的透射率进行修正,这些改进可以使复原图像的视觉效果更好。仿真实验证明改进算法可以实现预期效果。

关键词:  , 暗通道先验, 大气光亮度值A, k-means聚类, 自适应分割阈值T

Abstract:

Aiming at the problems of classical algorithm can’t obtain the accurate atmospheric brightness A and the restored image is easy to appear the color distortion
in the sky region, this paper proposes an improved algorithm which sets three sky thresholds to obtain the pixels in the sky region, then averages the brightness of these pixels
to eliminate the effect of a white object; Adaptive the threshold which is used to segment the sky region and non-sky region using k-means clustering algorithm. This improved
algorithm can obtain more accurate atmospheric brightness A, and it can segment the sky region more accurately, so it can rectify the transmittance of the sky region easily,
these improvements can make a better visual effect of restored image. Simulation results show that the improved algorithm can achieve the anticipated effect.

Key words:  dark channel prior, atmospheric brightness A, k-means clustering, adaptive threshold T