Single Image Dehazing using CNN

IEEE-xplore, 2023

Image dehazing is a rapid research area in the field of video surveillance and computer applications. Images are degraded by the scattering of atmospheric particles in the light, fog, and air particles in the atmosphere. In this paper, we are discussing conventional existing techniques such as the Dark Channel Prior method (DCP), Median Dark Channel Prior Method (MDCP), and Average Image Dehazing Model, but these methods suffer from a lot of computing complexity and even visual distortions like halos and over-saturation. To reduce the computational complexity and recover the original image from the hazy image Deep Learning and Convolutional Neural networks (CNN) are used. In this paper, we discussed various methods to reduce the haze in an image by using a feature/learning-based Convolutional Neural Network (CNN) such as Attention-based Transmission Estimation and Classification Fusion Network (ATECFN), and Encoder-Decoder architecture.

International Journal for Research in Applied Science and Engineering Technology IJRASET, 2020

Haze in images is due to natural environmental phenomena, which makes the image in a white shade noise. Haze removal is one of the most important research topics these days to due popularity of applications in real time surveillance from drones or any area under security. Both indoor and outdoor images are important for testing haze and its removal. Many image processing techniques are made by researchers to remove haze in a single image. Haze intensity can be calculated by a parameter known as perceptual fog density (PFD). It is important to analyze this parameter for all the techniques so as to get an idea of improvement. In this paper , a new approach is made by applying globally guided filtering technique with deep neural network. This proposed algorithm is implemented on MATLAB software and results are obtained by calculating the PFD in the existing and proposed technique. The four techniques are compared with each other. The techniques are global filtering (GIF), weighted global filtering(WGIF), Globally guided filtering(GGIF) and proposed technique i.e. Globally guided filtering with DNN (Deep Neural Network). In GIF, the fine structure of the image is generally not preserved and unrealistic image is obtained. In WGIF, the PFD obtained is highest. In GGIF, PFD is lower and Structure is not preserved, but in proposed algorithm, the PDF is minimum with fine structure, color intensity of the picture is of the best quality.

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2019

In this work we propose HR-Dehazer, a novel and accurate method for image dehazing. An encoder-decoder neural network is trained to learn a direct mapping between a hazy image and its respective clear version. We designed a special loss that forces the network to keep into account the semantics of the input image and to promote consistency among local structures. In addition, this loss makes the system more invariant to scale changes. Quantitative results on the recently released Dense-Haze dataset introduced for the NTIRE2019-Dehazing Challenge demonstrates the effectiveness of the proposed method. Furthermore, qualitative results on real data show that the described solution generalizes well to different never-seen scenarios.

2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2018

This paper proposes a novel approach to remove haze degradations in RGB images using a stacked conditional Generative Adversarial Network (GAN). It employs a triplet of GAN to remove the haze on each color channel independently. A multiple loss functions scheme, applied over a conditional probabilistic model, is proposed. The proposed GAN architecture learns to remove the haze, using as conditioned entrance, the images with haze from which the clear images will be obtained. Such formulation ensures a fast model training convergence and a homogeneous model generalization. Experiments showed that the proposed method generates high-quality clear images. 1

Computer Vision – ECCV 2020 Workshops, 2020

Learning to dehaze single hazy images, especially using a small training dataset is quite challenging. We propose a novel generative adversarial network architecture for this problem, namely back projected pyramid network (BPPNet), that gives good performance for a variety of challenging haze conditions, including dense haze and inhomogeneous haze. Our architecture incorporates learning of multiple levels of complexities while retaining spatial context through iterative blocks of UNets and structural information of multiple scales through a novel pyramidal convolution block. These blocks together for the generator and are amenable to learning through back projection. We have shown that our network can be trained without over-fitting using as few as 20 image pairs of hazy and non-hazy images. We report the state of the art performances on NTIRE 2018 homogeneous haze datasets for indoor and outdoor images, NTIRE 2019 denseHaze dataset, and NTIRE 2020 non-homogeneous haze dataset.

2022

A novel Encoder-Decoder Network with Guided Transmission Map (EDN-GTM) for single image dehazing scheme is proposed in this paper. The proposed EDN-GTM takes conventional RGB hazy image in conjunction with its transmission map estimated by adopting dark channel prior as the inputs of the network. The proposed EDN-GTM utilizes U-Net for image segmentation as the core network and utilizes various modifications including spatial pyramid pooling module and Swish activation to achieve state-of-the-art dehazing performance. Experiments on benchmark datasets show that the proposed EDN-GTM outperforms most of traditional and deep learning-based image dehazing schemes in terms of PSNR and SSIM metrics. The proposed EDN-GTM furthermore proves its applicability to object detection problems. Specifically, when applied to an image preprocessing tool for driving object detection, the proposed EDN-GTM can efficiently remove haze and significantly improve detection accuracy by 4.73% in terms of mAP...

2018 Colour and Visual Computing Symposium (CVCS), 2018

We compare a recent dehazing method based on deep learning, Dehazenet, with traditional state-of-the-art approaches, on benchmark data with reference. Dehazenet estimates the depth map from transmission factor on a single color image, which is used to inverse the Koschmieder model of imaging in the presence of haze. In this sense, the solution is still attached to the Koschmieder model. We demonstrate that the transmission is very well estimated by the network, but also that this method exhibits the same limitation than others due to the use of the same imaging model.

International Journal of Electrical and Computer Engineering (IJECE), 2025

The removal of noise from images holds great significance as clear and denoised images are vital for various applications. Recent research efforts have been concentrated on the dehazing of single images. While conventional methods and deep learning approaches have been employed for daytime images, learning-based techniques have shown impressive dehazing results, albeit often with increased complexity. This has led to the persistence of prior-based methods, despite their slightly lower performance. To address this issue, we propose a novel deep learning-based dehazing method utilizing a self-supervised convolutional neural network (CNN). This approach incorporates both the input hazy image and the dark channel prior. By leveraging an encoder, the combined information of the dark channel prior and haze image is encoded into a condensed latent representation. Subsequently, a decoder is employed to reconstruct the clean image using these latent features. Our experimental results demonstrate that our proposed algorithm significantly enhances image quality, as indicated by improved peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) values. We perform both quantitative and qualitative comparisons with recently published techniques, showcasing the efficacy of our approach.

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

This paper reviews the second NTIRE challenge on image dehazing (restoration of rich details in hazy image) with focus on proposed solutions and results. The training data consists from 55 hazy images (with dense haze generated in an indoor or outdoor environment) and their corresponding ground truth (haze-free) images of the same scene. The dense haze has been produced using a professional haze/fog generator that imitates the real conditions of haze scenes. The evaluation consists from the comparison of the dehazed images with the ground truth images. The dehazing process was learnable through provided pairs of haze-free and hazy train images. There were ∼ 270 registered participants and 23 teams competed in the final testing phase. They gauge the state-of-the-art in image dehazing.

Research Square (Research Square), 2023

Image haze removal has a significant involvement in numerous computer vision (CV) related uses. The primary purpose of the present article is to outline the current deep learning (based on CNN) processes used for single image dehazing. The previous problems with the haze removal methods based on multiple images are first addressed. Then, fundamental concepts of atmospheric scattering model and Convolutional Neural Network (CNN) are explained. The currently available single image dehazing approaches are divided into 3 groups: prior based, image fusion based and deep learning based approaches. Highlights and challenges of these dehazing techniques are discussed. The synthetic and real datasets utilized by different researchers in dehazing techniques are described along with implementation details. The paper also mentions performance metrics for evaluating image quality.