基于多尺度特征结合细节恢复的单幅图像去雾方法

为提高单幅图像去雾方法的准确性及其去雾结果的细节可见性,该文提出一种基于多尺度特征结合细节恢复的单幅图像去雾方法。首先,根据雾在图像中的分布特性及成像原理,设计多尺度特征提取模块及多尺度特征融合模块,从而有效提取有雾图像中与雾相关的多尺度特征并进行非线性加权融合。其次,构造基于所设计多尺度特征提取模块和多尺度特征融合模块的端到端去雾网络,并利用该网络获得初步去雾结果。再次,构造基于图像分块的细节恢复网络以提取细节信息。最后,将细节恢复网络提取出的细节信息与去雾网络得到的初步去雾结果融合得到最终清晰的去雾图像,实现对去雾后图像视觉效果的增强。实验结果表明,与已有代表性的图像去雾方法相比,所提方法能够对合成图像及真实图像中的雾进行有效去除,且去雾结果细节信息保留完整。     

Optimized contrast enhancement for real-time image and video dehazing

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.     

A joint cumulative distribution function and gradient fusion based method for dehazing of long shot hazy images

Hazy or foggy weather conditions significantly degrade the visual quality of an image in an outdoor environment. It also changes the color and reduces the contrast of an image. This paper introduces a novel single image dehazing technique to restore a hazy image without considering the physical model of haze formation. In order to find haze-free image, the proposed method does not require the transmission map and its costly refinement process. Since haze effect is dependent on the depth, it severely degrades the visibility of the objects located at a far distance. The objects close to the camera are unaffected. In this paper, we propose a fusion-based haze removal method based on the joint cumulative distribution function (JCDF) that treats faraway haze and nearby haze separately. The output images after the JCDF module, fused in the gradient domain to produce a haze-free image. The proposed method not only significantly enhances visibility but also preserves texture details. The proposed method is experimented and evaluated on a large set of challenging hazy images (large scene depth, night time, dense fog, etc.). Both qualitative and quantitative measures show that the performance of the proposed method is better than the state-of-the-art dehazing techniques.     

基于霾层学习的单幅图像去雾算法

针对传统去雾算法出现色彩失真、去雾不完全、出现光晕等现象,本文提出了一种基于霾层学习的卷积神经网络的单幅图像去雾算法.首先,依据大气散射物理模型进行理论推导,本文设计了一种能够直接学习和估计有雾图像和霾层图像之间的映射关系的网络模型.采用有雾图像作为输入,并输出有雾图像与无雾图像之间的残差图像,随后直接从有雾图像中去除此霾层图像,即可恢复出无雾图像.残差学习的引入,使得网络来直接估计初始霾层,利用相对大的学习率,减少计算量,加快收敛过程.再利用引导滤波进行细化,使得恢复出的无雾图像更接近真实场景.本文对不同雾浓度的有雾图片的去雾效果进行测试,并与当前主流深度学习去雾算法及其他经典算法进行对比.实验结果显示,本文设计的卷积神经网络模型在图像去雾的应用,不论在主观效果还是客观指标上,都有优势.     

基于迁移学习的三子网图像去雾方法

目前大部分图像去雾算法只在一种或几种均匀雾图数据集中有较好的表现,对于不同风格或非均匀雾图数据集去雾效果较差,同时算法在实际应用中会因模型泛化能力差导致模型场景受限。针对上述情况,该文提出一种基于迁移学习的卷积神经网络(CNN)用于解决去雾算法中非均匀雾图处理效果不佳和模型泛化能力差等问题。首先,该文使用ImageNet预训练的模型参数作为迁移学习模型的初始参数,以加速模型训练收敛速度。其次,主干网络模型由3个子网组成:残差特征子网络、局部特征提取子网络和整体特征提取子网络。3子网结合以保证模型可从整体和局部两个方面进行特征提取,在现实雾场景(浓雾、非均匀雾)中获得较好的去雾效果。该文在模型训练效率、去雾质量和雾图场景选择灵活性3个方面进行了研究和改进,为衡量模型性能,模型选择在去雾难度较大的非均匀雾图数据集NTIRE2020和NTIRE2021上进行定量与定性实验。实验结果证明3子网模型在图像主观和客观评价指标两个方面都取得了较好的效果。该文模型改善了算法泛化性能差和小数据集难以进行模型训练的问题,可将该文成果广泛应用于小规模数据集和多变场景图像的去雾工作中。     

DCA-CycleGAN: Unsupervised single image dehazing using Dark Channel Attention optimized CycleGAN

Single image dehazing has great significance in computer vision. In this paper, we propose a novel unsupervised Dark Channel Attention optimized CycleGAN (DCA-CycleGAN) to deal with the challenging scene with uneven and dense haze concentration. Firstly, the DCA-CycleGAN adopts the dark channel as input and then generate attention through a DCA subnetwork to handle the nonhomogeneous haze. Secondly, in addition to the conventional global discriminator, we also leverage two local discriminators to enhance the dehazing performance on the local dense haze, and a new local adversarial loss calculated strategy is been proposed. Specifically, the dehazing generator consists of two subnetworks: an auto-encoder and a dark channel attention subnetwork. The auto-encoder consists of an encoder, a feature transformation module, and a decoder. The dark channel attention subnetwork has the same structure as the encoder and the feature transformation module to ensure the same receptive field, which utilizes the dark channel to generate attention map and fine-tune the auto-encoder. Experimental results against several state-of-the-art methods demonstrate that our method can generate better visual effects, and is effective.     

Adaptive single image dehazing method based on support vector machine

A dehazing method often only shows good results when processing the image for a certain haze concentration. So an adaptive hazy image dehazing method based on SVM is proposed. The innovation points are as follows: Firstly, combining the characteristics of the degraded images of haze weather, the dark channel histogram and texture features of the input images are extracted to form the feature vectors. These are trained by supervised learning through SVM algorithm to realize automatic binary classification of images; Secondly, the defined dehazing methods are called to process the classified result as a hazy image and the same quality evaluation indexes are used to evaluate each image output by different dehazing methods. Then, it outputs the highest evaluation image after haze removal. Finally, the output image is classified again by SVM until the image reaches the clearest it can be. The experimental results show that the proposed algorithm exhibits good contrast, brightness and color saturation from the visual effect. Also the scene adaptability and robustness of the algorithm are improved.     

基于双注意力机制的雾天图像清晰化算法研究

针对传统去雾算法容易依赖先验知识以及恢复出来的清晰图像会产生颜色失真等问题,本文提出一种基于双注意力机制的雾天图像清晰化算法。首先将雾图输入编码器,经过下采样后得到特征图像;特征提取模块将多个特征提取基本块联结在一起,每个基本块由局部残差学习和特征注意模块组成,提高图像质量以及图像特征信息的利用率,增加网络训练的稳定性;然后通过通道注意力与多尺度空间注意力并行的结构处理特征图像,使得网络更加关注细节特征,提取更多关键信息,同时提高网络效率;最后将融合后的特征图像输入解码器中,经过多级映射,得到与输入大小匹配的雾密度图。实验结果表明,不论是对合成雾天图像或者真实雾天图像,本文算法能够高效地进行去雾处理,得到更自然的清晰图像。     

From synthetic to natural — single natural image dehazing deep networks using synthetic dataset domain randomization

Image dehazing methods aim to solve the problem of poor visibility in images due to haze. Techniques proposed for image dehazing in literature focus on image priors, haze lines or data driven statistical models. Variations of the classical methods relying on prior model or haze line model use no-reference image quality metrics to prove their dehazing performance. Recently developed deep learning models rely on huge amounts of hazy, haze-free pairs for training, and uses PSNR and SSIM like image reconstruction metrics to show their performance. These methods perform poorly on no-reference image quality assessments and also dehazes poorly at the depths of the image. These methods though can be optimized for memory usage and are faster. This work presents a deep learning model (Feature Fusion Attention Network) trained on a domain randomized synthetic dataset generated in simulation. The proposed model achieves the highest scores on blind image assessments through the gradient rationing technique for a deep learning-based approach by a significant margin. The images were evaluated on full-reference metrics as well and obtained favorable results. This approach also yields one of the highest edge sharpness obtained after dehazing. The training procedure adopted to obtain significant gains on real-world dehazing, without using any real-world data is also detailed in this paper.     

Generative adversarial network-based atmospheric scattering model for image dehazing

This paper presents a trainable Generative Adversarial Network (GAN)-based end-to-end system for image dehazing, which is named the DehazeGAN. DehazeGAN can be used for edge computing-based applications, such as roadside monitoring. It adopts two networks: one is generator (G), and the other is discriminator (D). The G adopts the U-Net architecture, whose layers are particularly designed to incorporate the atmospheric scattering model of image dehazing. By using a reformulated atmospheric scattering model, the weights of the generator network are initialized by the coarse transmission map, and the biases are adaptively adjusted by using the previous round’s trained weights. Since the details may be blurry after the fog is removed, the contrast loss is added to enhance the visibility actively. Aside from the typical GAN adversarial loss, the pixel-wise Mean Square Error (MSE) loss, the contrast loss and the dark channel loss are introduced into the generator loss function. Extensive experiments on benchmark images, the results of which are compared with those of several state-of-the-art methods, demonstrate that the proposed DehazeGAN performs better and is more effective.     

基于雾气特征与不等关系约束的去雾方法

针对现有去雾算法未充分考虑图像雾气信息、复原图像细节模糊等问题,提出一种新颖的反映图像雾信息分布的雾气特征图,并采用不等关系约束方法提高图像质量。首先,提取退化图像的极值通道以实现雾气信息的粗略估计,并通过L-1正则化对其进行优化从而得到雾气特征图。其次,提出一种基于雾气特征的初级大气光幕函数,通过对颜色通道和大气光幕作深入分析,利用均值不等式获得约束后的退化场景大气光幕。最后,利用雾气特征图对局部大气光进行改进,并基于大气散射模型实现图像去雾。将所提算法在真实雾图和合成数据集雾图上与其他经典方法进行比较分析,可以发现,所提算法在单幅图像去雾中展现了较好的性能,且在夜间雾图复原中更具优势。     

图像去雾过程中的噪声抑制方法

大气中微小颗粒（如雾、霾等）的散射作用会使户外场景拍摄的图像发生退化,造成图像质量下降。图像去雾可以提升图像对比度,增加场景能见度,校正颜色失真,改善视觉效果。但是图像去雾经常会出现明显的噪声放大现象,尤其是无穷远处的天空区域最为严重。针对这一问题,提出了一种去雾过程中的噪声抑制方法。以传输率图像为指导,采用滤波半径变化的双边滤波对雾天图像进行模糊。再计算新的传输率图像,代入雾天成像模型,得到去噪后复原图像。结合噪声评价方法,实验结果验证了该方法的噪声抑制效果。     

Variational optimization based single image dehazing

In this paper, we present a new approach for single image dehazing based on the proposed variational optimization. A hazy image captures the information about haze in terms of the transmission map and object details present in it. We propose to estimate the initial transmission map by performing the structure-aware smoothing of the hazy image. Further, we formulated a variational optimization for the estimation of final transmission, which refines the initial transmission of a hazy image. Atmospheric light can be considered to be constant throughout the scene for practical purposes. The uniform atmospheric light is computed from the dark channel of a hazy image. The exhaustive experimentation shows that the performance of the proposed method is comparable or better.     

基于光照估计的夜间图像去雾

相对白天雾天图像,夜晚雾天图像具有整体亮度低、光照不均匀、偏色等特点,因此去雾难度大。本文从夜间雾天成像规律出发,提出了基于光照估计的夜间图像去雾算法。针对光照不均匀问题,通过估计光照图来去除不均匀光照的影响;针对目前白天去雾算法假设不适用于夜晚图像问题,提出基于信息熵的传输图粗估计的方法;针对颜色失真问题,通过统计光源区域的颜色属性来进行颜色校正。实验结果表明,本文算法能够有效的去除不均匀光照影响,提高图像对比度,改善图像视觉效果。     

A modular architecture for high resolution image dehazing

Haze is an atmospheric phenomenon which diminishes visibility in outdoor images. Algorithms based on dark channel prior (DCP) and haze line prior are found to be effective for dehazing images. These two methods make use of the Laplacian matrix, which is computationally complex, memory intensive and slow, thus making it impossible to use them on high-resolution (large) images. Multiple strategies have been suggested in the literature to speed up dehazing process by avoiding the Laplacian matrix, but these methods compromise on the quality of dehazing. We propose an effective modular method which divides the input image into blocks and processes each block independently. This makes it possible to use our method for dehazing large images retaining Laplacian matting and thus ensuring the output image quality. This division results in the possibility of assuming local values of atmospheric light. We show that this approach results in better dehazing in the local regions. The effectiveness of the proposed modular architecture is tested also on a learning based method. The output of the modular method is compared with those of different state-of-the-art dehazing methods for multiple quality parameters. Toward this, we have created a dataset of hazy natural outdoor images of large size.     

基于改进CycleGAN的ISAR图像超分辨方法

图像超分辨是解决ISAR欺骗干扰中由于模型样本不完备导致难以对大带宽ISAR实现高逼真假目标模拟的重要手段。利用生成对抗网络（GAN）可通过端到端映射实现ISAR图像的超分辨,然而,当测试输入样本与训练输入样本分辨率差异较大时,超分辨图像中会出现伪散射点从而导致目标失真。考虑到循环生成对抗网络（CycleGAN）对输入样本差异适应性较好,本文提出了一种基于改进CycleGAN的ISAR欺骗干扰超分辨样本生成方法,分别从损失函数、优化过程、判别器结构三方面对CycleGAN网络结构进行改进,加快了网络的收敛速度,同时对于输入分辨率差异较大的ISAR图像泛化性能更好。利用暗室测量数据验证了所提方法的有效性,与GAN方法相比,对于训练输入样本分辨率差异较大的测试输入样本,生成的超分辨样本散射点位置与真实数据具有更好的匹配效果。     

Real-world image dehazing with improved joint enhancement and exposure fusion

In this work, a single image dehazing method that improves the haze removal capacity of the Joint Contrast Enhancement and Exposure Fusion (CEEF) method with Smoothing-Sharpening Image Filter (SSIF) is presented. In this method, the hazy image is first sharpened with SSIF to obtain a sharper image. In this way, the difference between haze and objects is amplified. Then, the AHE procedure in CEEF is replaced by CLAHE to obtain an enhanced CEEF. The enhanced CEEF is applied to the filtering result to obtain the final dehazed image. Observations demonstrate that the proposed method obtains enhanced results while reducing the amount of haze. The visual and quantitative comparisons between the proposed method and state-of-the-art dehazing methods show that the proposed method has better dehazing performance and has a 50% improvement in terms of the FADE metric compared to the closest result.     

基于深度学习的图像去雾算法研究进展

图像是信息的重要承载形式。雾霾的出现降低了图像采集设备采集到的图像质量,容易出现色彩暗淡、对比度和饱和度降低、细节信息丢失等问题,直接影响了有用信息的表达和利用。目前对图像去雾的研究多采用深度学习的方法,卷积神经网络代替了人工特征提取方式,取得了优于传统算法的去雾效果,但普遍存在着对真实世界雾霾图像和清晰图像对的依赖。无监督学习的方法带来了新的解决思路。从监督学习和无监督学习的角度对有代表性的深度学习图像去雾算法进行分类,归纳了常用的数据集、评价指标,概括分析了有影响力的去雾模型的核心思想,总结了各算法的优缺点和适用场景。针对目前工作存在的不足,探索了下一步研究的方向。     

Aethra-net: Single image and video dehazing using autoencoder

A fast and efficient video dehazing system with low computational complexity has a huge demand among drivers during hazy winter nights. There are only a few video dehazing models that exist in literature. Video dehazing requires the sequential extraction and processing of frames. The processed frames must be restored in the same sequence as the original video. However, the existing video dehazing algorithms suffer from color distortion due to the continuous processing of frames. They are not suitable for videos with dense haze. Furthermore, some dehazing systems require hardware, whereas the proposed model is completely software-based to reduce the computational costs. In this paper, an image and video dehazing system called Aethra-Net is developed. A gush enhancer-based autoencoder is modified to obtain the transmission map. The structure of gush enhancement module resembles the processing of light entering the human eye from different paths. The multiple blocks of Resnet-101 layers are employed to overcome vanishing gradient problem. The vessel enhancement filter is also incorporated to enhance the performance of the proposed system. The proposed model has a susceptibility to compute the dehazed images effectively. The proposed model is evaluated on various benchmark datasets and compared with the existing dehazing techniques. Experimental results reveal that the performance of Aethra-Net is found superior as compared to the existing dehazing models.     

基于多先验约束和一致性正则的半监督图像去雾算法

针对合成雾霾图像训练的去雾模型在真实场景中去雾效果不佳、对高层视觉任务性能提升不明显等问题,该文提出一种基于多先验约束和一致性正则的半监督图像去雾算法。该方法采用编码器-解码器网络结构,同时在合成雾霾图像与真实雾霾图像上学习去雾映射,并利用多种统计先验去雾结果作为真实雾霾图像参考真值进行半监督学习,同时通过多张真实雾霾图像的随机混合进行一致性正则约束,以消除多种先验去雾结果差异以及噪声干扰,提高图像去雾结果的视觉质量。实验对比结果表明,所提算法可比现有方法获得更好的真实场景去雾结果,并且能够显著提升高层视觉任务性能。