摘要
针对图像去雾领域缺乏有效雾浓度的检测方法,通过引入Naka-Rushton(NR)拟合函数,提出基于广义灰度差-比散点图先验的图像雾浓度检测算法。首先,提取不同条件下标准图像集灰度散点图的广义灰度差-比先验(generalized pixel difference ratio,GPDR);其次,引入符合先验约束的Naka-Rushton拟合函数,通过计算标准图像集拟合NR函数的参数(n,k),建立(n,k)与视场雾浓度对应的查找表;再次,采用回归分析法计算真实有雾图像拟合参数(n',k'),并计算标准参数(n,k)与真实拟合参数(n',k')间的综合相关系数,通过搜索(n,k)查找表2评定雾浓度等级。通过不同浓度有雾图像测试,证明算法测试结果符合浓度变化趋势:经过同场景不同浓度、不同场景不同浓度样本测试,算法测试结果与PM2.5相关系数达0.95,表明算法能够作为视场雾浓度等级评定;经过横向对比测试表明研究算法测试误差小于4.8%,可以用于视场雾浓度检测。
Addressing the limitations of fog concentration inspection in image defogging,an algorithm based on the scatterplot prior of the generalized pixel difference-ratio(GPDR)and the Naka-Rushton(NR)fitting function was proposed.First,the GPDR prior for gray scatterplots in standard foggy image sets across various scenes was extracted.Next,the NR function,constrained by the prior,was introduced,and a lookup table of parameters(n,k)corresponding to fog concentration levels was established by calculating the parameters(n,k)of NR function for standard image sets.Regression analysis was then used to calculate the parameters(n',k')for real foggy images,and the comprehensive correlation coefficient between(n,k)and(n',k')was calculated.Parameters(n,k)with correlation coefficients exceeding a set threshold were considered indicative of the fog concentration level.Simulations show that the algorithm accurately reflect changes in fog concentration across images with varying densities.Additionally,correlation coefficients between the algorithm’s results and PM2.5 measurements reached up to 0.95,both within the same and across different scenes.This shows that the algorithm can be effectively used for fog concentration rating in visual field.Horizontal comparison tests show that the inspection accuracy of the proposed algorithm can reach up to 4.8%,making it suitable for field fog concentration detection.
作者
温立民
王会峰
巨永锋
WEN Limin;WANG Huifeng;JU Yongfeng(School of Electronic&Engineering,Chang’an University,Xi’an 710064,P.R.China;Experimental Teaching Center of Electronics and Electronics,Chang’an University,Xi’an 710064,P.R.China)
出处
《重庆大学学报》
北大核心
2025年第4期115-126,共12页
Journal of Chongqing University
基金
国家自然科学基金(52172324)
陕西省交通厅重点项目(20-38T)。
