地球信息科学学报 2018 , 20 (2): 228-234 https://doi.org/10.12082/dqxxkx.2018.170366

遥感科学与应用技术

分割暗通道先验邻域的单幅图像去雾算法

黄黎红

莆田学院机电工程学院,莆田 351100

The Algorithm of Segmenting the Prior Neighborhood of Dark Channel in the Single Image Dehazing

HUANG Lihong^*^,

College of Mechanical & Electrical Engineering, Putian University, Putian 351100, China

通讯作者: *Corresponding author: HUANG Lihong, E-mail: 894209214@qq.com

收稿日期: 2017-08-7

修回日期: 2017-11-9

网络出版日期: 2018-03-02

基金资助: 国家自然科学基金项目（11172138）福建省自然科学基金项目（2012J05008）

作者简介:

作者简介：黄黎红（1971-）,女,硕士,教授,研究方向为光学测试、混合图像处理。E-mail: 894209214@qq.com

展开

摘要

利用暗原色先验进行单幅图像去雾时,需采取高计算复杂度的细化程序,否则其估计的传输率易在边界处造成光晕。对导致边界处产生光晕现象的原因进行分析时发现,计算复杂度高的细化程序在去除晕轮效应时去雾过度,且传统的基于暗原色先验的单幅去雾算法在明亮区域易造成色彩失真现象。由此在原来的透射率估计时,提出一种基于色调的简单而快速的邻域分割方法。首先将原始RGB图像转换到HSI色彩空间,在H（Hue）通道中,用邻域中的点与中心点的色调的差值绝对值,来判断该邻域内的点是否属于同一区域,只使用属于同一区域的像素点来计算该区域的暗原色值;再通过修正透射率值,来校正明亮区域的色彩失真。在图像复原时,在HSI色彩空间保留色调分量不变,仅对强度分量运用修改的暗原色值进行去雾,再进行非线性增强,最后对饱和度分量进行颜色补偿。实验表明,本文的去雾算法能够显著提高场景的视觉清晰度,而且不需要图像后续修补,并能获得更好的色彩视觉保真。

关键词： 单幅图像去雾 ; 暗原色先验 ; 透射率 ; HSI色彩空间 ; 亮区域校正

Abstract

A refinement program of high computational complexity is needed to dehaze an image by using dark channel prior. It will avoid haloes at boundaries which is related to the transmission rate. In analyzing halo phenomenon at boundaries, it is founded that highly computational complexity of refinement procedures usually dehaze excessively, and the traditional methods based on dark channel prior for a single image dehazing may cause the color distortion in bright regions. Therefore, a simple and fast neighborhood segmentation method based on the hue is proposed during estimation of original transmittance. Firstly, the source RGB images are converted to HIS color space, In H (Hue) channel, differences in neighborhood of point and center point of the tone of absolute value determine whether those pixels in the neighborhood belong to the same region. Only those pixels belonging to the same areas are used to calculate Dark Channel. Then, transmission value corrects the color of bright region. When the image is restored, hue component remains unchanged in HIS color space. Only the intensity component is defogged using the modified dark channel values. Then, the nonlinear enhancement is performed. Finally, the saturation component is compensated by the color. Experiments show that the proposed algorithm can significantly improve the visual clarity of scenes and get better color fidelity without subsequent image repairing.

Keywords： single image dehazing ; dark channel prior ; transmission ; color space for HSI ; bright area modification

PDF (1448KB) 元数据多维度评价相关文章收藏文章

本文引用格式导出 EndNote Ris Bibtex

黄黎红. 分割暗通道先验邻域的单幅图像去雾算法[J]. 地球信息科学学报, 2018, 20(2): 228-234 https://doi.org/10.12082/dqxxkx.2018.170366

HUANG Lihong. The Algorithm of Segmenting the Prior Neighborhood of Dark Channel in the Single Image Dehazing[J]. Journal of Geo-information Science, 2018, 20(2): 228-234 https://doi.org/10.12082/dqxxkx.2018.170366

1 引言

雾霾等环境常常会使户外图像的质量严重退化,使图像清晰度很低,给图像的后续处理、视频监控和目标识别等造成很大的困扰,甚至失效。以往通常采用不同天气条件下同一场景的多幅图像^[1]或3D几何模型^[2]来获取场景的深度信息以实现图像去雾。近几年,单幅图像去雾取得了很大的进展,主要通过一些统计先验知识或假设来估计场景透射率。Tan^[3]通过无雾天气条件下的图像的对比度总是比有雾条件下高的特点,提出最大化有雾图像的局部对比度,以达到去雾的目的,但是这种算法容易引起图像过饱和,且在深度不连续的地方易出现光晕效应;Kim等^[4]提出通过最优化图像对比度来快速复原图像,但是这种算法的估计不过精细;Fattal^[5]假设大气透射率和场景反射率不相关,通过独立成分分析（Independent Component Analysis,ICA）来复原场景的透射率和辐射率,取得了较好的去雾效果,然而这种算法对浓雾情况下失效;Tarel等^[6]利用中值滤波器来实时估计大气光幂,但是其在景深突变区域的复原效果不甚令人满意;He等^[7]利用晴朗天气下户外图像至少有一个颜色通道的像素值很低的统计思想,提出暗通道先验（Dark Channel Prior,DCP）去雾算法,得到较好的去雾效果。但是暗通道算法存在3个主要的弊端：① 在景物边缘处容易因为透射率的不恰当估计而出现光晕现象;② 在强度比较大的区域（如天空、马路或白色物体等）由于暗通道法不成立,会造成图像的过增强而引起色彩失真;③ 图像需要用软抠图法进行后期修补,增加了计算复杂度,从而限制了其应用范围,导致后面又出现许多改进算法,如分别用双边滤波器^[8]、中值滤波器^[9]、边缘保护滤波器^[10]、引导滤波器^[11]等进行后期修复,提高了速度。Gibsin等^[12]提出中值暗通道（Median Dark Channel Prior,MDCP）算法,该算法把DCP算法中对邻域求最小值操作改为求中值操作,使光晕现象减弱,但并不能完全消除。Gibson等^[13]提出用椭圆内最暗像素的平均值来实现图像去雾;Yu^[14]提出块对点的插补算法来估计透射率;Xin等^[15]用数学形态学的开运算处理暗原色图像,保护了图像的边缘;Irfan等^[16]针对明亮区域透射率求解有可能溢出允许值,而修改了暗通道算法。本文提出一种结合色调的改进后暗通道算法,利用色调来重新定义暗通道的邻域,克服暗通道法由于景物深度不连续造成的透射率误估计而出现的光晕,同时修正了明亮区域由于过增强而造成的色彩失真,且去雾后图像无需后续修补,降低了计算复杂度,实验结果表明去雾后的图像清晰、色彩自然不失真。

2 暗通道先验原理

McCartney于1975年提出了著名的大气散射模型^[17],即：

$\begin{matrix} I (x) = J (x) t (x) + A (1 - t (x)) \end{matrix}$ （1）

式中： $\begin{matrix} x \end{matrix}$ 代表像元点的空间位置; $\begin{matrix} I (x) \end{matrix}$ 表示含雾图像; $\begin{matrix} J (x) \end{matrix}$ 表示去雾后的图像; $\begin{matrix} t (x) \end{matrix}$ 表示像元点 $\begin{matrix} x \end{matrix}$ 处的透射率; $\begin{matrix} A \end{matrix}$ 为大气光,通常假设为全局常量。

He等^[7]的暗通道法得出在晴朗天气下的户外图像总有至少一个通道的值趋于0,即 $\begin{matrix} \min_{x \in Ω (x)} (\min_{c \in {r, g, b})} J^{c} (x)) \to 0 \end{matrix}$ 。由式（1）可得：

$\begin{matrix} t (x) = 1 - w \min_{y \in Ω (x)} (\min_{c \in \{r, g, b\}} \frac{I^{c} (y)}{A^{c}}) \end{matrix}$ （2）

式（2）添加一个常数 $\begin{matrix} w \end{matrix}$ （0< $\begin{matrix} w \end{matrix}$ ≤ 1,通常取0.95）是为了增强视觉效果,让空间保留一定的雾,看起来更有一定的透视感。

3 暗通道算法的改进

3.1 邻域范围的改进

暗通道法中（式（2））成立的前提是在以 $\begin{matrix} x \end{matrix}$ 为中心的某一邻域 $\begin{matrix} Ω (x) \end{matrix}$ (通常取15×15的块区域)内透射率恒定,但是透射率是场景深度的函数。如果邻域内包含不同景物,那么不同景物的深度可能不同。换言之,在景物边界处,由于深度不连续,势必导致透射率恒定不成立,这是产生光晕的主要原因。因此,克服光晕的重点在于如何保证邻域内只包含同一个物体,不能把深度不同的景物包含在内。虽然采用减小块尺寸（如取3×3块）可以减小光晕,但它会引起其他的问题：① 在更小的块区域里,暗通道法可能不再成立;② 引起过增强。

为了克服暗通道法在块区域内因为景物深度不一致引起的失真,文献[18]提出基于图像分割的方法,分割后的每个局部区域是由相同的物体构成,因此可以看成处于相同的深度。但是完全靠图像分割来去雾也会引起其他的问题：① 在分割后的每个局部区域里,具有相同的透射率并不总是正确的。例如,马路等平坦区域通常被分割到同一块区域里,但是它们的深度变化很大,不应该具有相同的透射率;② 在分割后较大的局部块区域内存在欠增强现象,而相反,在较小的局部块区域内存在过增强现象。

基于以上分析,把暗通道法中的平均分割和基于图像分割2种方法结合起来,取长补短,做到既可以用暗通道法的平均分割,又可以有图像分割后景深相同的特点,以克服以上述缺点。Bui等^[19]采取判断邻域内点的像素值与中心点像素值的相对误差是否小于一给定的阈值来重新分割邻域,但是用此方法来确定重新分割后的邻域内是否包含同一物体并不准确。本文认为首先雾不会改变景物的色调,其次同一物体的色调一般相同,因此用色调来判断邻域内的点是否属于同一物体,更符合理论实际。具体为：把输入图像转换到HSI色彩空间,在暗通道邻域 $\begin{matrix} Ω (x) \end{matrix}$ （15×15的块区域）内,把与邻域中心点 $\begin{matrix} x \end{matrix}$ 处色调相近的区域归为新的邻域 $\begin{matrix} Ω^{S} (x) \end{matrix}$ （其中 $\begin{matrix} Ω^{S} (x) \subseteq Ω (x) \end{matrix}$ ）,即：

$\begin{matrix} if |\frac{H (y) - H (x)}{H (x)}| < ε ，则 y \in Ω^{S} (x) \end{matrix}$ （3）

式中： $\begin{matrix} H (x) \end{matrix}$ 代表邻域 $\begin{matrix} Ω (x) \end{matrix}$ 中心点 $\begin{matrix} x \end{matrix}$ 处的色调分量; $\begin{matrix} y \end{matrix}$ 为邻域 $\begin{matrix} Ω (x) \end{matrix}$ 内的点。 $\begin{matrix} H \end{matrix}$ 的取值范围是0-1,本实验中 $\begin{matrix} ε \end{matrix}$ 取0.08。故式（2）修改为：

$\begin{matrix} t (x) = 1 - w \min_{y \in Ω^{S} (x)} (\min_{c \in \{r, g, b\}} \frac{I^{c} (y)}{A^{c}}), Ω^{S} (x) \subseteq Ω (x) \end{matrix}$ （4）

3.2 明亮区域透射率溢出现象的修正

由式（1）可得：

$\begin{matrix} J (x) = \frac{I (x) - A}{t (x)} + A \end{matrix}$ （5）

根据式（5）可做出 $\begin{matrix} J \end{matrix}$ 与 $\begin{matrix} I \end{matrix}$ 的函数关系曲线图（设t=0.8, A=225）。

由图1可看出,当 $\begin{matrix} I \end{matrix}$ 取值超出某一范围的值时, $\begin{matrix} J \end{matrix}$ 可能溢出（0,255）的范围。特别是在明亮区域,当 $\begin{matrix} J > 255 \end{matrix}$ 时,通常的做法是令其值为255,这就造成了失真。为此,对式（1）两边分别求2次最大值操作,得：

$\begin{matrix} \max_{x \in Ω (x)} (\max_{c \in \{r, g, b\}} I^{c} (x)) = \max_{x \in Ω (x)} (\max_{c \in \{r, g, b\}} J^{c} (x)) t (x) + A^{c} (1 - t (x)) \end{matrix}$ （6）

当 $\begin{matrix} I^{c} > A^{c} \end{matrix}$ 时,由于 $\begin{matrix} \max_{x \in Ω (x)} (\max_{c \in \{r, g, b\}} J^{c} (x)) \leq 255 \end{matrix}$ ,由式（6）得：

$\begin{matrix} t (x) \geq \max_{x \in Ω (x)} (\max_{c \in \{r, g, b\}} \frac{I^{c} (x) - A^{c}}{255 - A^{c}}) \end{matrix}$ （7）

对于明亮区域,R、G、B通道的最小值与最大值接近,故式（7）可改为：

（8）

最终的透射率应满足：

$\begin{matrix} t (x) = \max \{1 - w \min_{y \in Ω^{S} (x)} (\min_{c \in \{r, g, b\}} \frac{I^{c} (y)}{A^{c}}), \\ \min_{y \in Ω^{S} (x)} (\min_{c \in \{r, g, b\}} \frac{I^{c} (y) - A^{c}}{255 - A^{c}}), τ_{0}\} \end{matrix}$ （9）

为了不使分母为0,需限制一个下限 $\begin{matrix} τ_{0} \end{matrix}$ （通常取0.1）。暗通道法与本文算法在明亮区域（ $\begin{matrix} τ_{0} \end{matrix}$ 在0.1附近,设 $\begin{matrix} A \end{matrix}$ 取225）透射率值的比较（图2）可知,经过本算法修改后的透射率值在明亮区域得到抑制。

显示原图| 下载原图ZIP| 生成PPT

图1 J与I的关系（t取0.8, A取225）

Fig. 1 Relationship between J and I (t=0.8, A=225)

显示原图| 下载原图ZIP| 生成PPT

图2 本文算法和He算法的透射率比较

Fig. 2 The transmission comparison between He’s algorithm and the proposed algorithm

3.3 估计全局大气光

在图像去雾中,全局大气光 $\begin{matrix} A \end{matrix}$ 的求解正确与否非常重要,它能影响整幅图像的亮度和色彩。若估计的 $\begin{matrix} A \end{matrix}$ 值过大,会引起去雾后的图像太暗;相反,过小会引起去雾后的图像太亮。而有些 $\begin{matrix} A \end{matrix}$ 值的估计是错误的,如Tan^[3]用亮度最大值作为大气光值,它的处理前提是太阳光的影响可以忽略,且图像中没有比大气光更亮的白色物体。为了克服这些问题,He等^[7]选取暗原色值的第0.1%个最大值的像素点,映射到原图像中,取这些像素点的最大值作为全局大气光 $\begin{matrix} A \end{matrix}$ ,但是图像中白色物体占整幅图像的比例不定,所以这种方法仍然不够准确。Kim等^[20]把图像的灰度级进行四叉树分割,把分割后最亮区域的平均灰度作为大气光,但这种方法仍然易把马路、河流等白色区域误判断为天空区域;Wang等^[21]认为天空区域一定位于图像的上方而改进了四叉树分割算法,取得了较好的效果,但是此算法较复杂。本文采取简单的方法,认为：① 首先天空区域必定位于图像的上方;② 天空区域强度较大,接近白色;③天空区域的亮度梯度比较平坦。而在HSI色彩空间中,强度（Intensity）分量 $\begin{matrix} V \end{matrix}$ （取值范围为0-1）接近1时,即为白色区域。故只需把 $\begin{matrix} V \end{matrix}$ 分量进行大尺度平均分割（如采用45×45块）,选取位于图像上方前1/4内、块区域 $\begin{matrix} V \end{matrix}$ 分量梯度最小、平均值最大（或较大）的强度平均值作为大气光值。

3.4 饱和度复原

暗通道算法分别在R、G、B通道对像元点进行处理,由于图像色彩的相关性,R、G、B没有同比增强,会造成图像色彩失真。由于雾对图像的色调没有影响,但会减小图像的饱和度,增强图像的亮度,因此把图像转换到HSI色彩空间后,假设其色调分量、饱和度分量、强度分量分别为 $\begin{matrix} H \end{matrix}$ 、 $\begin{matrix} S \end{matrix}$ 、 $\begin{matrix} V \end{matrix}$ ,则保持色调分量不变,运用式（8）、（5）对强度分量进行去雾处理,最后对饱和度分量进行复原补偿,这样既减少了计算量,又保持了图像的色彩,复原公式为^[22,23]：

$\begin{matrix} V_{out} (x) = \frac{V (x) - A}{\max (t (x), t_{0})} + A \end{matrix}$ （10）

（11）

式中： $\begin{matrix} V_{out} \end{matrix}$ 、 $\begin{matrix} V \end{matrix}$ 、 $\begin{matrix} S_{out} \end{matrix}$ 、 $\begin{matrix} S \end{matrix}$ 分别为去雾前后的强度分量和饱和度分量。

4 图像复原后增强

He算法不仅在明亮区域易造成颜色失真,而且经He算法复原后的图像在暗区更暗,从而造成暗区细节丢失,因此复原后的图像应该适当增大暗区强度,并抑制亮区强度,使图像的局部细节更加清晰、丰富。由于通过中心/围绕函数与亮度分量的卷积后可获得局部均值,则局部增强细节可由式（12）非线性变换计算得到。

$\begin{matrix} V_{out}' (x, y) = [V_{out} (x, y) - F (x, y) * V_{out} {(x, y)]}^{γ} \end{matrix}$ （12）

$\begin{matrix} F (x, y) = K e^{- (x^{2} + y^{2}) / σ^{2}} \end{matrix}$ （13）

其中, $\begin{matrix} K \end{matrix}$ 由归一化函数决定：

$\begin{matrix} \iint F (x, y) dxdy = 1 \end{matrix}$ （14）

式中： $\begin{matrix} σ \end{matrix}$ 为高斯函数的尺度参数; $\begin{matrix} γ \end{matrix}$ 为非线性调节系数。为了增强图像暗区域的细节,应适当减小 $\begin{matrix} σ \end{matrix}$ 、 $\begin{matrix} γ \end{matrix}$ 的值（本实验 $\begin{matrix} σ \end{matrix}$ 取2-5, $\begin{matrix} γ \end{matrix}$ 取0.7-0.8）。本算法流程如图3所示。

显示原图| 下载原图ZIP| 生成PPT

图3 本算法流程图

Fig. 3 The flow chart of the algorithm

5 实验结果及性能分析

通过实验对比来验证本文算法的性能。图4是He^[7]算法与本文算法的去雾效果比较。从图4第一行可看出,He算法去雾后,图中的人物雕像边缘出现明显的白色光晕（尤其是左边手臂下方）,且由于R、G、B三分量没有同比增强,图像色彩出现严重偏离;而本算法去雾后图像没有光晕现象,且色彩与原图相比,基本不失真。从图4第二行可看出,图中的蓝色天空和白云区域出现了明显的过增强;且图4的暗区域比原图更暗（如大山和树木）,造成细节的丢失;而结本算法去雾后天空等明亮区域不失真,且暗处细节变亮,细节更加丰富,而且图像的色彩与原图比较,基本上比较接近,颜色更自然。由图5的本文算法去雾前后的效果对比可看出,图像整体清晰度明显增强,同时暗区域得到增强,细节突出,色彩自然逼真。

显示原图| 下载原图ZIP| 生成PPT

图4 He算法和本文算法去雾结果比较

Fig. 4 Comparison between He’s algorithm and the proposed algorithm

显示原图| 下载原图ZIP| 生成PPT

图5 本算法去雾效果

Fig. 5 Image dehazing results obtained by the proposed method

由图6的Fattal、He、Tan算法与本文算法的处理结果比较可以看出,He、Tan算法在天空处出现了明显的颜色失真,且暗处细节丢失比较严重,而经本文处理后的图像不仅清晰度大,颜色基本上不失真,且在天空区域的细节复原也比较理想。

根据图像色调的还原程度取决于原图像与复原后图像的直方图形状是否大体一致,可用2幅图像的直方图相似度 $\begin{matrix} d \end{matrix}$ （式（15））来度量图像色彩的保真程度^[24]。

式中： $\begin{matrix} h \end{matrix}$ 、 $\begin{matrix} \overset{′}{h} \end{matrix}$ 分别为去雾前后图像的灰度值; $\begin{matrix} \overset{̅}{h} \end{matrix}$ 、 $\begin{matrix} \overset{′}{\overset{̅}{h}} \end{matrix}$ 分别为 $\begin{matrix} h \end{matrix}$ 、 $\begin{matrix} \overset{′}{h} \end{matrix}$ 的均值;k取值为0-255。 $\begin{matrix} d \end{matrix}$ 越大,说明色调的保真能力越强。

为此,本文对图6中4种算法处理结果进行评价（熵、平均梯度、色调还原度 $\begin{matrix} d \end{matrix}$ 以及运行时间）,如表1所示（图像大小为576像元×768像元）。本文程序是在第六代Intel Core i5 CPU,4G内存、显卡520,Windows7操作系统的PC机上运行。

显示原图| 下载原图ZIP| 生成PPT

图6 4种算法和处理效果比较

Fig. 6 The comparison of the algorithm in this study with other methods

表1 目标比较算法的客观质量评估

Tab. 1 Objective quality assessments of all comparison algorithms

	源图像	方法
	源图像	Fattal	He	Tan	本文方法
熵	18.1632	25.0137	21.9613	10.4437	28.0772
平均梯度	2.1267	2.3036	2.3714	2.4727	2.5156
色调还原度		0.7325	0.7162	0.5172	0.8063
运行时间/s		5.9413	58.2260	581.1476	2.8762

新窗口打开

6 结论

暗通道法在求透射率时是假设平均分割的各方块（邻域）内景物深度恒定,但这种假设在深度不连续区域失效。因此,本文通过判断邻域内各点的色调与邻域中心点的色调的差值是否在阈值内,来重新划分邻域,从而确保新的邻域内只包含同一物体,即具有相同的深度,以防止透射率误估计而出现的光晕;并修正了天空等明亮区域过增强而造成的失真,最后复原由于雾的稀释作用而造成的饱和度下降情况。本文算法无需图像后期修补。实验表明,经过本算法处理后的图像清晰度高、且图像颜色较自然,算法运算速度较快。

The authors have declared that no competing interests exist.

参考文献

原文顺序

文献年度倒序

文中引用次数倒序

被引期刊影响因子

[1]	Narasimhan S G, Nayar S K. Vison and atmosphere [J]. International Journal of Computer Vision, 2002,48:233-254. https://doi.org/10.1023/A:1016328200723 URL [本文引用: 1]
[2]	Johannes K, Boris N, Billy C, et al. Deep photo:Model-based photograph enhancement and viewing [J]. ACM Transactions on Graphics, 2008,27(5):116. https://doi.org/10.1145/1457515.1409069 URL [本文引用: 1] 摘要 In this paper, we introduce a novel system for browsing, enhancing, and manipulating casual outdoor photographs by combining them with already existing georeferenced digital terrain and urban models. A simple interactive registration process is used to align a photograph with such a model. Once the photograph and the model have been registered, an abundance of information, such as depth, texture, and GIS data, becomes immediately available to our system. This information, in turn, enables a variety of operations, ranging from dehazing and relighting the photograph, to novel view synthesis, and overlaying with geographic information. We describe the implementation of a number of these applications and discuss possible extensions. Our results show that augmenting photographs with already available 3D models of the world supports a wide variety of new ways for us to experience and interact with our everyday snapshots.
[3]	Tan R T.Visibility in bad weather from a single image[C]//Proceedings of the 26th IEEE Conference on Computer Vision and Pattern Recognition.Los Alamitos: IEEE Computer Society Press, 2008. [本文引用: 2]
[4]	Kim J H, Jang W D, Sim J Y, et al. Optimized contrast enhancement for real-time image and video dehazing [J]. Journal of Visual Communication and Image Representation, 2013,24(3):410-425. https://doi.org/10.1016/j.jvcir.2013.02.004 URL [本文引用: 1] 摘要 A fast and optimized dehazing algorithm for hazy images and videos is proposed in this work. Based on the observation that a hazy image exhibits low contrast in general, we restore the hazy image by enhancing its contrast. However, the overcompensation of the degraded contrast may truncate pixel values and cause information loss. Therefore, we formulate a cost function that consists of the contrast term and the information loss term. By minimizing the cost function, the proposed algorithm enhances the contrast and preserves the information optimally. Moreover, we extend the static image dehazing algorithm to real-time video dehazing. We reduce flickering artifacts in a dehazed video sequence by making transmission values temporally coherent. Experimental results show that the proposed algorithm effectively removes haze and is sufficiently fast for real-time dehazing applications.
[5]	Fattal R. Single image dehazing [J]. ACM Transactions on Graphics, 2008,27(3):547-555. [本文引用: 1]
[6]	Tarel J P, Hautiere N.Fast visibility restoration from a single color or gray level image[C]//Proceedings of the 12th IEEE International Conference on Computer Vision.Los Alamitos: IEEE Computer Society Press, 2009:2201-2208. [本文引用: 1]
[7]	He K M, Sun J, Tang X O. Single image haze removal using dark channel prior [J]. IEEE Transactions on Pattern Analysis Machine Interlligence, 2011,33(12):2341-2353. https://doi.org/10.1109/TPAMI.2010.168 URL [本文引用: 4]
[8]	Yu J,Li D, Liao Q. Physics-based fast single image fog removal [J]. Acta Automatica Sinica, 2011,37(2):143-149. https://doi.org/10.3724/SP.J.1004.2011.00143 URL [本文引用: 1]
[9]	Gibson K B, Vo D T, Nguyen T Q. An investigation of dehazing effects on image and video coding [J]. IEEE Transactions on Image Processing, 2012,21(2):662-673. https://doi.org/10.1109/TIP.2011.2166968 URL PMID: 21896391 [本文引用: 1] 摘要 This paper makes an investigation of the dehazing effects on image and video coding for surveillance systems. The goal is to achieve good dehazed images and videos at the receiver while sustaining low bitrates (using compression) in the transmission pipeline. At first, this paper proposes a novel method for single-image dehazing, which is used for the investigation. It operates at a faster speed than current methods and can avoid halo effects by using the median operation. We then consider the dehazing effects in compression by investigating the coding artifacts and motion estimation in cases of applying any dehazing method before or after compression. We conclude that better dehazing performance with fewer artifacts and better coding efficiency is achieved when the dehazing is applied before compression. Simulations for Joint Photographers Expert Group images in addition to subjective and objective tests with H.264 compressed sequences validate our conclusion.
[10]	Shiau Y H, Yang H Y, Chen P Y, et al. Hardware implementation of a fast and efficient haze remval method [J].IEEE Transactions on Circuits and Systems for Video Technology, 2013,23(8):1369-1374. https://doi.org/10.1109/TCSVT.2013.2243650 URL [本文引用: 1] 摘要 In this letter, a fast and efficient haze removal method is presented. We employ an extremum approximate method to extract the atmospheric light and propose a contour preserving estimation to obtain the transmission by using edge-preserving and mean filters alternately. Our method can efficiently avoid the halo artifact generated in the recovered image. To meet the requirement of real-time applications, an 11-stage pipelined hardware architecture for our haze removal method is presented. It can achieve 200 MHz with 12.8 K gate counts by using TSMC 0.13- 渭m technology. Simulation results indicate that our design can obtain comparable results with the least execution time compared to previous algorithms and is suitable for low-cost high-performance hardware implementation for haze removal.
[11]	He K M, Sun J, Tang X O. Guided image filtering [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013,35(6):1397-1409. https://doi.org/10.1109/TPAMI.2012.213 URL [本文引用: 1]
[12]	Gibson K B, Vo D T, Nguyen T Q. An investigation of dehazing effects on image and video coding [J]. IEEE Transactions on Image Processing, 2012,21(2):662-673. https://doi.org/10.1109/TIP.2011.2166968 URL PMID: 21896391 [本文引用: 1] 摘要 This paper makes an investigation of the dehazing effects on image and video coding for surveillance systems. The goal is to achieve good dehazed images and videos at the receiver while sustaining low bitrates (using compression) in the transmission pipeline. At first, this paper proposes a novel method for single-image dehazing, which is used for the investigation. It operates at a faster speed than current methods and can avoid halo effects by using the median operation. We then consider the dehazing effects in compression by investigating the coding artifacts and motion estimation in cases of applying any dehazing method before or after compression. We conclude that better dehazing performance with fewer artifacts and better coding efficiency is achieved when the dehazing is applied before compression. Simulations for Joint Photographers Expert Group images in addition to subjective and objective tests with H.264 compressed sequences validate our conclusion.
[13]	Gibson K B, Nguyen T Q. An analysis of single image defogging methods using a color ellipsoid framework [J]. Eurasip Journal on Image and Video Processing, 2013,2013(1):1-14. https://doi.org/10.1186/1687-5281-2013-1 URL [本文引用: 1] 摘要 This article presents a new approach for gait-based gender recognition using depth cameras, that can run in real time. The main contribution of this study is a new fast feature extraction strategy that uses the 3D point cloud obtained from the frames in a gait cycle. For each frame, these points are aligned according to their centroid and grouped. After that, they are projected into their PCA plane, obtaining a representation of the cycle particularly robust against view changes. Then, final discriminative features are computed by first making a histogram of the projected points and then using linear discriminant analysis. To test the method we have used the DGait database, which is currently the only publicly available database for gait analysis that includes depth information. We have performed experiments on manually labeled cycles and over whole video sequences, and the results show that our method improves the accuracy significantly, compared with state-of-the-art systems which do not use depth information. Furthermore, our approach is insensitive to illumination changes, given that it discards the RGB information. That makes the method especially suitable for real applications, as illustrated in the last part of the experiments section.
[14]	Yu T, Riaz I, Piao J, et al. Real-time single image dehazing using block-pixel interpolation and adaptive dark channel prior [J]. IET Image Processing, 2015,9(9):725-734. https://doi.org/10.1049/iet-ipr.2015.0087 URL [本文引用: 1] 摘要 The authors propose a novel and efficient method for single image dehazing. To accelerate the transmission estimation process, a block-to-pixel interpolation method is used for fine dark channel computation, in which the block-level dark channel is first computed, and then the fine pixel-level dark channel is obtained by a weighted voting of the block-level dark channel to preserve edges and smooth out texture noise. This technique can be used for a direct transmission map generation without a computationally expensive refinement step. Since the dark channel prior (DCP) is not valid in bright (sky) regions, they propose an adaptive DCP modelled by a Gaussian curve that produces a more natural recovered image of the sky and other bright regions. In addition, a scaling method for transmission map computation is proposed to further accelerate the dehazing method. Through experiments, they show that the proposed adaptive block-to-pixel technique is about 30 times faster and produces improved recovered images than the well-known state-of-the-art DCP approach.
[15]	Liu X, Zhang H, Tang Y Y, et al. Scene-adaptive single image dehazing via opening dark channel model [J]. IET image process, 2016,10(11):877-884. https://doi.org/10.1049/iet-ipr.2016.0138 URL [本文引用: 1] 摘要 Many traditional dark channel prior based haze removal schemes often suffer from the colour distortion and generate halo artefacts in the remote scenes. To tackle these issues, the authors present an efficient scene-adaptive single image dehazing approach via opening dark channel model (ODCM). First, the authors detect the image depth information and separate it into close view and distant view. Then, an ODCM is proposed to optimise the whole atmospheric veil, in which the values of close view are regularised by a minimum channel image while the distant parts are estimated by an appropriate lower constant. Accordingly, the transmission map can be further optimised by guide filter and smoothed by domain transform filter. Finally, the haze degraded image can be well restored by the atmosphere scattering model. The extensive experiments have shown that the proposed image dehazing approach has significantly increased the perceptual visibility of the scene and achieved a better colour fidelity visually.
[16]	Irfan R, Xue F, Hyunchul Shin. Single image dehazing with bright object handling [J]. IET computer Viaion, 2016,10(8):817-827. https://doi.org/10.1049/iet-cvi.2015.0451 URL [本文引用: 1] 摘要 This study addresses the shortcomings of the dark channel prior (DCP). The authors propose a new and efficient method for transmission estimation with bright-object handling capability. Based on the intensity value of a bright surface, they categorise DCP failures into two types: (i) obvious failure: occurs on surfaces that are brighter than ambient light. They show that, for these surfaces, altering the transmission value proportional to the brightness is better than the thresholding strategy; (ii) non-obvious failure: occurs on surfaces that are brighter than the neighbourhood average haziness value. Based on the observation that the transmission of a surface is loosely connected to its neighbours, the local average haziness value is used to recompute the transmission of such surfaces. This twofold strategy produces a better estimate of block and pixel-level haze thickness than DCP. To reduce haloes, a reliability map of block-level haze is generated. Then, via reliability-guided fusion of block- and pixel-level haze values, a high-quality refined transmission is obtained. Experimental results show that the authors' method competes well with state-of-the-art methods in typical benchmark images while outperforming these methods in more challenging scenarios. The authors' proposed reliability-guided fusion technique is about 60 times faster than other well-known DCP-based approaches.
[17]	McCartney E J. Optics of atmosphere:Scattering by molecules and particles [M]. New York: John Wiley and Sons, 1976:23-32. [本文引用: 1]
[18]	黄黎红. 单幅图像的去雾新算法 [J].光子学报,2011,40(9):1419-1422. URL [本文引用: 1] 摘要提出了一种基于单幅图像的去雾新算法.首先把图像归一化后从RGB彩色空间转换到HSI彩色空间，对色调分量运用四叉树分割法进行分割图像;分割后图像的每一局部小方块可以认为具有相同的场景深度，从而可以对每一局部小方块估计出空气光.然后再对亮度分量运用雾天图像光学模型，从雾的物理特性上去除雾对图像的影响.最后再对图像的饱和度分量进行校正，得到复原后的图像.该算法的主要优点是速度快，且不仅可以应用于彩色图像，也可以适用于灰度图像.通过该算法与其它几种算法的实验结果进行分析和比较，表明该算法能有效恢复出清晰图像. [ Huang L H. A novel algorithm for single image dehazing [J]. Acta Photonica Sinica, 2011,40(9):1419-1422. ] URL [本文引用: 1] 摘要提出了一种基于单幅图像的去雾新算法.首先把图像归一化后从RGB彩色空间转换到HSI彩色空间，对色调分量运用四叉树分割法进行分割图像;分割后图像的每一局部小方块可以认为具有相同的场景深度，从而可以对每一局部小方块估计出空气光.然后再对亮度分量运用雾天图像光学模型，从雾的物理特性上去除雾对图像的影响.最后再对图像的饱和度分量进行校正，得到复原后的图像.该算法的主要优点是速度快，且不仅可以应用于彩色图像，也可以适用于灰度图像.通过该算法与其它几种算法的实验结果进行分析和比较，表明该算法能有效恢复出清晰图像.
[19]	Bui T M, Tran H N, Kim W, et al. Segmenting dark prior in single image dehazing [J]. Electronics Letters, 2014,50(7):516-518. https://doi.org/10.1049/el.2013.3652 URL [本文引用: 1] 摘要 In image dehazing, the existing transmission estimators bring out the halo artefact at boundaries unless they adopt a refinement procedure with a high computational complexity. How the existing transmission estimation methods suffer from the halo artefact at the boundaries are analysed and it is found that the elaborate, high computational refinement procedure to remove the halo effect are excessive for dehazing. On the basis of the above-mentioned analysis and finding, a simple segmentation logic was embedded in an existing transmission estimator, which is sufficiently accurate for dehazing. The experiment verifies that the proposed method significantly reduces the halo artefact without requiring any refinement procedure.
[20]	Kim J H, Jang W D, Sim J Y, et al. Optimized contrast enhancement for real-time image and video dehazing [J]. Journal of Visual Communication and Image Representation, 2013,24(3):410-425. https://doi.org/10.1016/j.jvcir.2013.02.004 URL [本文引用: 1] 摘要 A fast and optimized dehazing algorithm for hazy images and videos is proposed in this work. Based on the observation that a hazy image exhibits low contrast in general, we restore the hazy image by enhancing its contrast. However, the overcompensation of the degraded contrast may truncate pixel values and cause information loss. Therefore, we formulate a cost function that consists of the contrast term and the information loss term. By minimizing the cost function, the proposed algorithm enhances the contrast and preserves the information optimally. Moreover, we extend the static image dehazing algorithm to real-time video dehazing. We reduce flickering artifacts in a dehazed video sequence by making transmission values temporally coherent. Experimental results show that the proposed algorithm effectively removes haze and is sufficiently fast for real-time dehazing applications.
[21]	Wang W C, Yuan X H, Wu X J, et al. Fast image dehazing method based on linear transformation [J]. IEEE Transactions on Multimedia, 2017,19(6):1142-1155. https://doi.org/10.1109/TMM.2017.2652069 URL [本文引用: 1] 摘要 Images captured in hazy or foggy weather conditions are seriously degraded by the scattering of atmospheric particles, which directly influences the performance of outdoor computer vision systems. In this paper, a fast algorithm for single image dehazing is proposed based on linear transformation by assuming that a linear relationship exists in the minimum channel between the hazy image and the haze-free image. First, the principle of linear transformation is analyzed. Accordingly, the method of estimating a medium transmission map is detailed and the weakening strategies are introduced to solve the problem of the brightest areas of distortion. To accurately estimate the atmospheric light, an additional channel method is proposed based on quad-tree subdivision. In this method, average grays and gradients in the region are employed as assessment criteria. Finally, the haze-free image is obtained using the atmospheric scattering model. Numerous experimental results show that this algorithm can clearly and naturally recover the image, especially at the edges of sudden changes in the depth of field. It can, thus, achieve a good effect for single image dehazing. Furthermore, the algorithmic time complexity is a linear function of the image size. This has obvious advantages in running time by guaranteeing a balance between the running speed and the processing effect.
[22]	黄黎红. 色彩空间中的单幅图像自适应去雾算法 [J]. 计算机辅助设计与图形学学报, 2015,27(8):1506-1511. https://doi.org/10.3969/j.issn.1003-9775.2015.08.019 URL [本文引用: 1] 摘要为解决传统基于暗原色先验的去雾算法在景物边界处出现的白边现象,以及在天空、白云等明亮区域出现的颜色失真的问题,提出一种基于色彩空间的快速去雾算法.首先把暗原色先验值求解中对邻域求最小值操作改为求中值操作;然后根据RGB色彩空间中像素颜色深度变化正好与场景深度变化相对应的关系,引入像素最大值与最小值之差的加权来调整暗原色先验值;再根据阈值来区分是否是明亮区域,修改明亮区域透射率估计的不足;最后在HIS色彩空间复原强度分量和饱和度值,并用直方图拉伸增强强度分量.文中算法无需图像后期修补,能自适应地修改暗原色先验值,实验结果表明了该算法的可行性和有效性. [ Huang L H. Adaptive defogging algorithm of single image in the color space [J].Journal of Computer-Aided Design ＆ Computer Graphics, 2015,27(8):1506-1511. ] https://doi.org/10.3969/j.issn.1003-9775.2015.08.019 URL [本文引用: 1] 摘要为解决传统基于暗原色先验的去雾算法在景物边界处出现的白边现象,以及在天空、白云等明亮区域出现的颜色失真的问题,提出一种基于色彩空间的快速去雾算法.首先把暗原色先验值求解中对邻域求最小值操作改为求中值操作;然后根据RGB色彩空间中像素颜色深度变化正好与场景深度变化相对应的关系,引入像素最大值与最小值之差的加权来调整暗原色先验值;再根据阈值来区分是否是明亮区域,修改明亮区域透射率估计的不足;最后在HIS色彩空间复原强度分量和饱和度值,并用直方图拉伸增强强度分量.文中算法无需图像后期修补,能自适应地修改暗原色先验值,实验结果表明了该算法的可行性和有效性.
[23]	黄素霞. 单幅图像中值暗通道先验去雾的改进算法 [J].光电子·激光,2015,26(8):1611-1617. URL [本文引用: 1] 摘要为解决传统基于暗通道先验（DCP）的单幅去雾算法在景物边界处出现的白边现象及高计算复杂度,在中值DCP（MDCP）法的基础上提出一种改进算法。首先通过判断RGB三颜色通道的最小值与中值暗原色的差值绝对值的大小判断DCP法或MDCP法是否成立：当DCP法成立时,取较大窗口的暗原色值作为最终的暗原色值;当DCP法不成立而MDCP法成立时,取两种窗口中中值暗原色值较小的值作为最终的暗原色值;当DCP法和MDCP法都不成立时,取较小窗口的中值暗原色值作为最终的暗原色值。然后在HSI色彩空间保留色调分量不变,仅对强度分量运用修改的暗原色值进行去雾。最后,对饱和度分量进行颜色补偿。本文算法无需图像后期修补。实验结果表明了本文改进算法的可行性和有效性。 [ Huang S X. Improved algorithm for the haze removing of the single image using median dark channel prior [J]. Journal of Optoelectronics·Laser, 2015,26(8):1611-1617. ] URL [本文引用: 1] 摘要为解决传统基于暗通道先验（DCP）的单幅去雾算法在景物边界处出现的白边现象及高计算复杂度,在中值DCP（MDCP）法的基础上提出一种改进算法。首先通过判断RGB三颜色通道的最小值与中值暗原色的差值绝对值的大小判断DCP法或MDCP法是否成立：当DCP法成立时,取较大窗口的暗原色值作为最终的暗原色值;当DCP法不成立而MDCP法成立时,取两种窗口中中值暗原色值较小的值作为最终的暗原色值;当DCP法和MDCP法都不成立时,取较小窗口的中值暗原色值作为最终的暗原色值。然后在HSI色彩空间保留色调分量不变,仅对强度分量运用修改的暗原色值进行去雾。最后,对饱和度分量进行颜色补偿。本文算法无需图像后期修补。实验结果表明了本文改进算法的可行性和有效性。
[24]	李大鹏,禹晶,肖创柏. 图像去雾的无参考客观质量评测方法 [J].中国图像图形学报,2011,16(9):1753-1757. https://doi.org/10.11834/jig.20110928 URL [本文引用: 1] 摘要许多学者对图像去雾进行了深入研究，但对去雾算法的客观评测方法却寥寥无几。本文根据图像去雾经常出现的Halo效应，色调偏移等问题，采用Canny算子和亮通道来检测图像的有效边缘强度，使用直方图相似性来度量色彩还原能力，并结合图像结构信息对去雾图像的质量进行客观评测。实验结果表明本文方法能够有效的对去雾算法进行比较与评价。 [ Li D P, Yu J, Xiao C B. No-reference quality assessment method for defogged images [J]. Journal of Image and Graphics, 2011,16(9):1753-1757. ] https://doi.org/10.11834/jig.20110928 URL [本文引用: 1] 摘要许多学者对图像去雾进行了深入研究，但对去雾算法的客观评测方法却寥寥无几。本文根据图像去雾经常出现的Halo效应，色调偏移等问题，采用Canny算子和亮通道来检测图像的有效边缘强度，使用直方图相似性来度量色彩还原能力，并结合图像结构信息对去雾图像的质量进行客观评测。实验结果表明本文方法能够有效的对去雾算法进行比较与评价。

Vison and atmosphere

2002

... 雾霾等环境常常会使户外图像的质量严重退化,使图像清晰度很低,给图像的后续处理、视频监控和目标识别等造成很大的困扰,甚至失效.以往通常采用不同天气条件下同一场景的多幅图像^[1]或3D几何模型^[2]来获取场景的深度信息以实现图像去雾.近几年,单幅图像去雾取得了很大的进展,主要通过一些统计先验知识或假设来估计场景透射率.Tan^[3]通过无雾天气条件下的图像的对比度总是比有雾条件下高的特点,提出最大化有雾图像的局部对比度,以达到去雾的目的,但是这种算法容易引起图像过饱和,且在深度不连续的地方易出现光晕效应;Kim等^[4]提出通过最优化图像对比度来快速复原图像,但是这种算法的估计不过精细;Fattal^[5]假设大气透射率和场景反射率不相关,通过独立成分分析（Independent Component Analysis,ICA）来复原场景的透射率和辐射率,取得了较好的去雾效果,然而这种算法对浓雾情况下失效;Tarel等^[6]利用中值滤波器来实时估计大气光幂,但是其在景深突变区域的复原效果不甚令人满意;He等^[7]利用晴朗天气下户外图像至少有一个颜色通道的像素值很低的统计思想,提出暗通道先验（Dark Channel Prior,DCP）去雾算法,得到较好的去雾效果.但是暗通道算法存在3个主要的弊端：① 在景物边缘处容易因为透射率的不恰当估计而出现光晕现象;② 在强度比较大的区域（如天空、马路或白色物体等）由于暗通道法不成立,会造成图像的过增强而引起色彩失真;③ 图像需要用软抠图法进行后期修补,增加了计算复杂度,从而限制了其应用范围,导致后面又出现许多改进算法,如分别用双边滤波器^[8]、中值滤波器^[9]、边缘保护滤波器^[10]、引导滤波器^[11]等进行后期修复,提高了速度.Gibsin等^[12]提出中值暗通道（Median Dark Channel Prior,MDCP）算法,该算法把DCP算法中对邻域求最小值操作改为求中值操作,使光晕现象减弱,但并不能完全消除.Gibson等^[13]提出用椭圆内最暗像素的平均值来实现图像去雾;Yu^[14]提出块对点的插补算法来估计透射率;Xin等^[15]用数学形态学的开运算处理暗原色图像,保护了图像的边缘;Irfan等^[16]针对明亮区域透射率求解有可能溢出允许值,而修改了暗通道算法.本文提出一种结合色调的改进后暗通道算法,利用色调来重新定义暗通道的邻域,克服暗通道法由于景物深度不连续造成的透射率误估计而出现的光晕,同时修正了明亮区域由于过增强而造成的色彩失真,且去雾后图像无需后续修补,降低了计算复杂度,实验结果表明去雾后的图像清晰、色彩自然不失真. ...

Deep photo:Model-based photograph enhancement and viewing

2008

... 在图像去雾中,全局大气光

\begin{matrix} A \end{matrix}

的求解正确与否非常重要,它能影响整幅图像的亮度和色彩.若估计的

\begin{matrix} A \end{matrix}

值过大,会引起去雾后的图像太暗;相反,过小会引起去雾后的图像太亮.而有些

\begin{matrix} A \end{matrix}

值的估计是错误的,如Tan^[3]用亮度最大值作为大气光值,它的处理前提是太阳光的影响可以忽略,且图像中没有比大气光更亮的白色物体.为了克服这些问题,He等^[7]选取暗原色值的第0.1%个最大值的像素点,映射到原图像中,取这些像素点的最大值作为全局大气光

\begin{matrix} A \end{matrix}

,但是图像中白色物体占整幅图像的比例不定,所以这种方法仍然不够准确.Kim等^[20]把图像的灰度级进行四叉树分割,把分割后最亮区域的平均灰度作为大气光,但这种方法仍然易把马路、河流等白色区域误判断为天空区域;Wang等^[21]认为天空区域一定位于图像的上方而改进了四叉树分割算法,取得了较好的效果,但是此算法较复杂.本文采取简单的方法,认为：① 首先天空区域必定位于图像的上方;② 天空区域强度较大,接近白色;③天空区域的亮度梯度比较平坦.而在HSI色彩空间中,强度（Intensity）分量

\begin{matrix} V \end{matrix}

（取值范围为0-1）接近1时,即为白色区域.故只需把

\begin{matrix} V \end{matrix}

分量进行大尺度平均分割（如采用45×45块）,选取位于图像上方前1/4内、块区域

\begin{matrix} V \end{matrix}

分量梯度最小、平均值最大（或较大）的强度平均值作为大气光值. ...

Optimized contrast enhancement for real-time image and video dehazing

2013

Single image dehazing

2008

2009

Single image haze removal using dark channel prior

2011

... He等^[7]的暗通道法得出在晴朗天气下的户外图像总有至少一个通道的值趋于0,即

\begin{matrix} \min_{x \in Ω (x)} (\min_{c \in {r, g, b})} J^{c} (x)) \to 0 \end{matrix}

.由式（1）可得： ...

... 在图像去雾中,全局大气光

\begin{matrix} A \end{matrix}

的求解正确与否非常重要,它能影响整幅图像的亮度和色彩.若估计的

\begin{matrix} A \end{matrix}

值过大,会引起去雾后的图像太暗;相反,过小会引起去雾后的图像太亮.而有些

\begin{matrix} A \end{matrix}

\begin{matrix} A \end{matrix}

\begin{matrix} V \end{matrix}

（取值范围为0-1）接近1时,即为白色区域.故只需把

\begin{matrix} V \end{matrix}

分量进行大尺度平均分割（如采用45×45块）,选取位于图像上方前1/4内、块区域

\begin{matrix} V \end{matrix}

分量梯度最小、平均值最大（或较大）的强度平均值作为大气光值. ...

... 通过实验对比来验证本文算法的性能.图4是He^[7]算法与本文算法的去雾效果比较.从图4第一行可看出,He算法去雾后,图中的人物雕像边缘出现明显的白色光晕（尤其是左边手臂下方）,且由于R、G、B三分量没有同比增强,图像色彩出现严重偏离;而本算法去雾后图像没有光晕现象,且色彩与原图相比,基本不失真.从图4第二行可看出,图中的蓝色天空和白云区域出现了明显的过增强;且图4的暗区域比原图更暗（如大山和树木）,造成细节的丢失;而结本算法去雾后天空等明亮区域不失真,且暗处细节变亮,细节更加丰富,而且图像的色彩与原图比较,基本上比较接近,颜色更自然.由图5的本文算法去雾前后的效果对比可看出,图像整体清晰度明显增强,同时暗区域得到增强,细节突出,色彩自然逼真. ...

Physics-based fast single image fog removal

2011

An investigation of dehazing effects on image and video coding

2012

Hardware implementation of a fast and efficient haze remval method

2013

Guided image filtering

2013

An investigation of dehazing effects on image and video coding

2012

An analysis of single image defogging methods using a color ellipsoid framework

2013

Real-time single image dehazing using block-pixel interpolation and adaptive dark channel prior

2015

Scene-adaptive single image dehazing via opening dark channel model

2016

Single image dehazing with bright object handling

2016

Optics of atmosphere:Scattering by molecules and particles

1976

... McCartney于1975年提出了著名的大气散射模型^[17],即： ...

单幅图像的去雾新算法

2011

... 为了克服暗通道法在块区域内因为景物深度不一致引起的失真,文献[18]提出基于图像分割的方法,分割后的每个局部区域是由相同的物体构成,因此可以看成处于相同的深度.但是完全靠图像分割来去雾也会引起其他的问题：① 在分割后的每个局部区域里,具有相同的透射率并不总是正确的.例如,马路等平坦区域通常被分割到同一块区域里,但是它们的深度变化很大,不应该具有相同的透射率;② 在分割后较大的局部块区域内存在欠增强现象,而相反,在较小的局部块区域内存在过增强现象. ...

单幅图像的去雾新算法

2011

Segmenting dark prior in single image dehazing

2014

... 基于以上分析,把暗通道法中的平均分割和基于图像分割2种方法结合起来,取长补短,做到既可以用暗通道法的平均分割,又可以有图像分割后景深相同的特点,以克服以上述缺点.Bui等^[19]采取判断邻域内点的像素值与中心点像素值的相对误差是否小于一给定的阈值来重新分割邻域,但是用此方法来确定重新分割后的邻域内是否包含同一物体并不准确.本文认为首先雾不会改变景物的色调,其次同一物体的色调一般相同,因此用色调来判断邻域内的点是否属于同一物体,更符合理论实际.具体为：把输入图像转换到HSI色彩空间,在暗通道邻域

\begin{matrix} Ω (x) \end{matrix}

（15×15的块区域）内,把与邻域中心点

\begin{matrix} x \end{matrix}

处色调相近的区域归为新的邻域

\begin{matrix} Ω^{S} (x) \end{matrix}

（其中

\begin{matrix} Ω^{S} (x) \subseteq Ω (x) \end{matrix}

）,即： ...

Optimized contrast enhancement for real-time image and video dehazing

2013

... 在图像去雾中,全局大气光

\begin{matrix} A \end{matrix}

的求解正确与否非常重要,它能影响整幅图像的亮度和色彩.若估计的

\begin{matrix} A \end{matrix}

值过大,会引起去雾后的图像太暗;相反,过小会引起去雾后的图像太亮.而有些

\begin{matrix} A \end{matrix}

\begin{matrix} A \end{matrix}

\begin{matrix} V \end{matrix}

（取值范围为0-1）接近1时,即为白色区域.故只需把

\begin{matrix} V \end{matrix}

分量进行大尺度平均分割（如采用45×45块）,选取位于图像上方前1/4内、块区域

\begin{matrix} V \end{matrix}

分量梯度最小、平均值最大（或较大）的强度平均值作为大气光值. ...

Fast image dehazing method based on linear transformation

2017

... 在图像去雾中,全局大气光

\begin{matrix} A \end{matrix}

的求解正确与否非常重要,它能影响整幅图像的亮度和色彩.若估计的

\begin{matrix} A \end{matrix}

值过大,会引起去雾后的图像太暗;相反,过小会引起去雾后的图像太亮.而有些

\begin{matrix} A \end{matrix}

\begin{matrix} A \end{matrix}

\begin{matrix} V \end{matrix}

（取值范围为0-1）接近1时,即为白色区域.故只需把

\begin{matrix} V \end{matrix}

分量进行大尺度平均分割（如采用45×45块）,选取位于图像上方前1/4内、块区域

\begin{matrix} V \end{matrix}

分量梯度最小、平均值最大（或较大）的强度平均值作为大气光值. ...

色彩空间中的单幅图像自适应去雾算法

2015

... 暗通道算法分别在R、G、B通道对像元点进行处理,由于图像色彩的相关性,R、G、B没有同比增强,会造成图像色彩失真.由于雾对图像的色调没有影响,但会减小图像的饱和度,增强图像的亮度,因此把图像转换到HSI色彩空间后,假设其色调分量、饱和度分量、强度分量分别为

\begin{matrix} H \end{matrix}

、

\begin{matrix} S \end{matrix}

、

\begin{matrix} V \end{matrix}

,则保持色调分量不变,运用式（8）、（5）对强度分量进行去雾处理,最后对饱和度分量进行复原补偿,这样既减少了计算量,又保持了图像的色彩,复原公式为^[22,23]： ...

色彩空间中的单幅图像自适应去雾算法

2015

\begin{matrix} H \end{matrix}

、

\begin{matrix} S \end{matrix}

、

\begin{matrix} V \end{matrix}

单幅图像中值暗通道先验去雾的改进算法

2015

\begin{matrix} H \end{matrix}

、

\begin{matrix} S \end{matrix}

、

\begin{matrix} V \end{matrix}

单幅图像中值暗通道先验去雾的改进算法

2015

\begin{matrix} H \end{matrix}

、

\begin{matrix} S \end{matrix}

、

\begin{matrix} V \end{matrix}

图像去雾的无参考客观质量评测方法

2011

... 根据图像色调的还原程度取决于原图像与复原后图像的直方图形状是否大体一致,可用2幅图像的直方图相似度

\begin{matrix} d \end{matrix}

（式（15））来度量图像色彩的保真程度^[24]. ...

图像去雾的无参考客观质量评测方法

2011

... 根据图像色调的还原程度取决于原图像与复原后图像的直方图形状是否大体一致,可用2幅图像的直方图相似度

\begin{matrix} d \end{matrix}

（式（15））来度量图像色彩的保真程度^[24]. ...

〈

〉

分割暗通道先验邻域的单幅图像去雾算法

The Algorithm of Segmenting the Prior Neighborhood of Dark Channel in the Single Image Dehazing

1 引言

2 暗通道先验原理

3 暗通道算法的改进

3.1 邻域范围的改进

3.2 明亮区域透射率溢出现象的修正

3.3 估计全局大气光

3.4 饱和度复原

4 图像复原后增强

5 实验结果及性能分析

6 结论

参考文献 文献选项 原文顺序 文献年度倒序 文中引用次数倒序 被引期刊影响因子

参考文献

原文顺序

文献年度倒序

文中引用次数倒序

被引期刊影响因子