The invention discloses a method to fog night image statistical characteristics and brightness estimation based on including: night fog image reversal have reversed image; calculation of inverted image local atmospheric light with color filter, and through the guidance of optimization; initial calculation of three channel and three channel transmission rate of the rough estimation of transmittance the use of bright, channel inversion image transmission on the three channel estimate is corrected, and optimize the transmission through guidance filter: image restoration after reversal to night with color shift to fog image using light and atmospheric transmittance with solving the local optimization and the local color; the grey word color correction final night to fog image. The brightness and contrast of the night image obtained by the invention not only can effectively restore the image to the fog image, but also the effective correction of night image color, improve visual effect, while preserving the image details, and greatly reduce the computational complexity.
【技术实现步骤摘要】
基于统计特性和亮度估计的夜晚图像去雾方法
本专利技术涉及一种计算机图像处理方法,尤其涉及一种夜晚图像去雾方法。
技术介绍
夜晚有雾环境下拍摄图像时,会导致图像整体灰度值和对比度降低并且丧失大量的细节信息,难以识别感兴趣的区域,给视频监控、室外目标识别与追踪、遥感成像等带来很大困难。因此,夜晚图像去雾问题在计算机视觉应用领域和数字图像处理领域亟待解决。现有的夜晚图像去雾方法较少,主要有Pei[1]提出的基于暗原色先验和颜色转换的夜晚图像去雾算法,Zhang[2]提出的基于新模型的去雾算法以及Li[3]的基于相对平滑约束的层次分解去雾算法等。这些算法的去雾主框架仍然是基于暗原色先验,但是由于夜晚有雾图像特殊的成像环境,暗原色先验在夜晚环境下并不适用,因此这些算法复原出的图像整体偏暗,并且存在着不同程度的颜色失真,在图像光源处晕轮效应明显,去雾不完全,而且计算复杂。[参考文献][1]PeiSC,LeeTY.Nighttimehazeremovalusingcolortransferpre-processinganddarkchannelprior[A].ProceedingsoftheIEEEInternationalConferenceonImageProcessing[C].Orlando:IEEEComputerSocietyPress,2012,957-960。[2]ZhangJ,CaoY,WangZ.Nighttimehazeremovalbasedonanewimagingmodel[A].ProceedingsoftheIEEEInternatio ...
【技术保护点】
一种基于统计特性和亮度估计的夜晚图像去雾方法,其特征在于,步骤如下:步骤1、输入的夜晚有雾图像为图像I(x),将图像I(x)反转得反转图像
【技术特征摘要】
1.一种基于统计特性和亮度估计的夜晚图像去雾方法,其特征在于,步骤如下:步骤1、输入的夜晚有雾图像为图像I(x),将图像I(x)反转得反转图像式(1)中,c∈{r,g,b};步骤2、计算反转图像局部带色偏的大气光rL(x)AL(x),并通过指导性滤波进行优化:式(2)中,Ω(x)为像素x的局部邻域,Ω(y)为邻域y的局部邻域,GF表示指导性滤波;步骤3、计算反转图像的三通道的初始透射率taL(x)和三通道的粗估透射率tbL(x):步骤4、利用反转图像的亮通道对三通道的粗估透射率tbL(x)进行校正,并通过指导性滤波进行优化:tL(x)=GF(eA_lighttbL(x))(6)式(5)和式(6)中:A_light为反转图像的亮通道,tL(x)为对三通道的粗估透射率tbL(x)校正后的优化透...
【专利技术属性】
技术研发人员:杨爱萍,杨苏辉,王南,赵美琪,
申请(专利权)人:天津大学,
类型:发明
国别省市:天津,12
还没有人留言评论。发表了对其他浏览者有用的留言会获得科技券。