Hazy Images Research Articles

HLA-HOD: Joint High-Low Adaptation for Object Detection in Hazy Weather Conditions

Object detection remains challenging in hazy weather conditions due to the poor visibility of captured images. There are currently two types of detectors capable of adapting to varying weather conditions: (i) low-level adaptation methods that combine one detector with an additional dehazing network and (ii) high-level adaptation methods that explore various kinds of domain adaptation knowledge. However, neither of these approaches can achieve desirable performance due to their inherent limitations. We raise an intriguing question—if combining both low-level adaptation and high-level adaptation, can improve the generalization ability of a detector in hazy weather conditions? To answer it, we propose a Joint High-Low Adaptation Object Detection paradigm (HLA-HOD) in hazy weather conditions. By combining both low-level adaptation and high-level adaptation, HLA-HOD achieves superior performance on hazy images without requiring ground-truth bounding boxes or clean images. Extensive experiments demonstrate that our method outperforms state-of-the-art low-level and high-level adaptation methods by a large margin both quantitatively and qualitatively.

Self-supervised zero-shot dehazing network based on dark channel prior

Most learning-based methods previously used in image dehazing employ a supervised learning strategy, which is time-consuming and requires a large-scale dataset. However, large-scale datasets are difficult to obtain. Here, we propose a self-supervised zero-shot dehazing network (SZDNet) based on dark channel prior, which uses a hazy image generated from the output dehazed image as a pseudo-label to supervise the optimization process of the network. Additionally, we use a novel multichannel quad-tree algorithm to estimate atmospheric light values, which is more accurate than previous methods. Furthermore, the sum of the cosine distance and the mean squared error between the pseudo-label and the input image is applied as a loss function to enhance the quality of the dehazed image. The most significant advantage of the SZDNet is that it does not require a large dataset for training before performing the dehazing task. Extensive testing shows promising performances of the proposed method in both qualitative and quantitative evaluations when compared with state-of-the-art methods.Graphical

An Aerial Image Dehazing Algorithm Using a Prior-Based Dense Attentive Network

<p>To address the problem that the acquired images tend to degrade in clarity and fidelity in aerial hazy conditions to the extent that the target is difficult to detect, this paper proposes an aerial image dehazing algorithm using a prior-based dense attentive network. The network is based on dense blocks and attention blocks with an encoder-decoder structure, which can directly learn the mapping between the input image and the corresponding haze-free image without relying on the traditional atmospheric scattering model. In addition, to better handle inhomogeneous hazy images, the initial fuzzy density map is first extracted from the original hazy images and then used as a common input to the network together with the original hazy images. Finally, this paper synthesizes a large -scale aerial image dehazing dataset containing two subsets of uniform and non -uniform images. The experimental results and data analysis show that the proposed method exhibits better performance of dehazing with other algorithms on both synthetic and real images.</p>

Multi-Purpose Oriented Single Nighttime Image Haze Removal Based on Unified Variational Retinex Model

Under the nighttime haze environment, the quality of acquired images will be deteriorated significantly owing to the influences of multiple adverse degradation factors. In this paper, we develop a multi-purpose oriented haze removal framework focusing on nighttime hazy images. First, we construct a nonlinear model based on the classic Retinex theory to formulate multiple adverse degradations of a nighttime hazy image. Then, a novel variational Retinex model is presented to simultaneously estimate a smoothed illumination component and a detail-revealed reflectance component and predict the noise map from a pre-processed nighttime hazy image in a unified manner. Specifically, an <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${\ell _{0}}$ </tex-math></inline-formula> norm is imposed on the reflectance to reveal the structural details and we make use of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\ell _{1}$ </tex-math></inline-formula> norm to constrain the piece-wise smoothness of the illumination and apply <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${\ell _{2}}$ </tex-math></inline-formula> norm to enforce the total intensity of the noise map. Afterwards, the haze in the illumination component is removed based on prior-based dehazing method and the contrast of the reflectance component is improved in the gradient domain. Finally, we combine the dehazed illumination and the improved reflectance to generate the haze-free image. Experiments show that our proposed framework performs better than famous nighttime image dehazing methods both in visual effects and objective comparisons. In addition, the proposed framework can also be applicable to other types of degraded images.

Retinex decomposition and fusion dehazing network

Image dehazing has been widely used in vision-based fields, such as detection, segmentation, traffic monitoring, and automated vehicle system. However, most of the existing end-to-end dehazing networks are fully data driven without physical constraints or prior information guidance, leading to difficulties in exploring latent structures and statistical characteristics of hazy images. We propose a novel Retinex decomposition-fusion dehazing network consisting of a dual-branch decomposition module and a fusion optimization module. Different from existing solutions, we decompose the clear images in the commonly used RESIDE dataset based on Retinex theory to construct the clear illumination map and reflection map datasets to drive the network training, equivalently imposing reasonable constraints on the network and achieving impressive dehazing performances. The dual-branch decomposition module is developed to estimate the illumination map and the reflection map, respectively. The illumination map mainly contains the global features of the image, whereas the reflection map reflects the inherent color properties of the image and contains rich details, with which we explore the latent structures and statistical characteristics of hazy images. In addition, the dual-branch structure avoids the error accumulation and information cancellation existing in current methods. Subsequently, the estimated illumination map and reflection map are fused and refined via the fusion optimization module to access the dehazed image. Experiments show that the proposed network has better generalization and visual effects than existing fully data-driven methods and can be applied successfully to real-world scenarios.

Boundary-constrained robust regularization for single image dehazing

Generally, for single image dehazing, regularization-based schemes improve the initial transmission map iteratively by using a guidance map as a structural prior. We conducted experiments on a large number of hazy images and observed that a constrained transmission map affects the quality of the recovered image. However, regularization-based methods do not constrain the transmission map to its physically valid range during the iterative process. It degrades its robustness to outliers, and consequently, deteriorates the quality of the recovered image. In addition, conventional methods fuse the structural information of the guidance and initial transmission map without considering any structural differences between them. To address these issues, in this paper, we present a robust regularization scheme that constraints the transmission map during its enhancement. In the proposed scheme, a nonconvex energy function is constructed that leverages the mutual structural information of the guidance and transmission map. The nonconvex problem is solved by a majorize-minimization algorithm, and the intermediate transmission maps are constrained through the appropriate lower and upper bounds. The retrieved transmission map has better edge-preserving properties, and ultimately, results in a high-quality haze-free image that has faithful colors and fine details. The proposed scheme is tested on benchmark datasets and results are evaluated through quantitative metrics. The comparative analysis has revealed the effectiveness of the proposed scheme.

MFID-Net: Multi-scaled feature-fused image dehazing via dynamic weights

Methods based on convolutional neural networks have achieved excellent performance in the image dehazing task. Unfortunately, most of the dehazing methods that exist suffer from loss of detail in the convolution and activation operations and failure to consider the effects of superimposing different intensities of haze, such as under-exposed and over-exposed images. To address these issues, we propose a dynamic dehazing convolution (DDC) based on attentional weight calculation and dynamic weight fusion and a dynamic dehazing activation (DDA) based on the input global context encoding function to address the problem of detail loss. And we propose a multi-scaled feature-fused image dehazing network (MFID-Net) based on DDC and DDA to address the effects of haze superposition. We also design a loss function based on the physical model with dynamic weights. Extensive experimental results demonstrate that the proposed MFID-Net performs favorably against the state-of-the-art algorithms on the hazy dataset while improving further on hazy images with large differences in haze concentration, and producing satisfactory dehazing results. The code is available at https://github.com/awhitewhale/MFID-Net.

IoT-Fog-enabled robotics-based robust classification of hazy and normal season agricultural images for weed detection

Abstract The mechanization of farming is currently the most pressing problem facing humanity and a burgeoning academic field. Over the last decade, there has been an explosion of Internet of Things (IoT) application growth in agriculture. Agricultural robotics is bringing about a new era of farming because they are growing more intelligent, recognizing causes of variation on the farm, consuming fewer resources, and optimizing their efficiency to more flexible jobs. The purpose of this article is to construct an IoT-Fog computing equipped robotic system for the categorization of weeds and soy plants during both the hazy season and the normal season. The used dataset in this article included four classes: soil, soybean, grass, and weeds. A two-dimensional Convolutional Neural Network (2D-CNN)-based deep learning (DL) approach was implemented for data image classification with dataset of height and width of 150 × 150 and of three channels. The overall proposed system is considered an IoT-connected robotic device that is capable of applying classification through the Internet-connected server. The reliability of the device is also enhanced as it is enabled with edge-based Fog computing. Hence, the proposed robotic system is capable of applying DL classification through IoT as well as Fog computing architecture. The analysis of the proposed system was conducted in steps including training and testing of CNN for classification, validation of normal images, validation of hazy images, application of dehazing technique, and at the end validation of dehazed images. The training and validation parameters ensure 97% accuracy in classifying weeds and crops in a hazy environment. Finally, it concludes that applying the dehazing technique before identifying soy crops in adverse weather will help achieve a higher classification score.

A Novel Parameter Adaptive Dual Channel MSPCNN Based Single Image Dehazing for Intelligent Transportation Systems

Visibility issues in intelligent transportation systems are exacerbated by bad weather conditions such as fog and haze. It has been observed from recent studies that major road accidents have occurred in the world due to low visibility and inclement weather conditions. Single image dehazing attempts to restore a haze-free image from an unconstrained hazy image. We proposed a dehazing method by cascading two models utilizing a novel parameter-adaptive dual-channel modified simplified pulse coupled neural network (PA-DC-MSPCNN). The first model uses a new color channel for removing haze from images. The second model is the improved brightness preserving model (I-GIHE), which retains the brightness of the image while improving the gradient strength. To integrate the results from these two models and provide a pleasing haze-free image, a PA-DC-MSPCNN-based fusion is used. Furthermore, the proposed approach is deployed on a Xilinx Zynq SoC by exploiting the recently released PYNQ platform. The dehazing system runs on a PYNQ-Z2 all-programmable SoC platform, where it will input the camera feed through the FPGA unit and carry out the dehazing algorithm in the ARM core. This configuration has allowed reaching real-time processing speed for image dehazing. The results of dehazing are analyzed using both synthetic and real-world hazy images. Synthetic hazy images are acquired from the O-HAZE, I-HAZE, SOTS, and FRIDA datasets, while real-world hazy images are taken from the RailSem19, E-TUVD dataset, and the internet. For evaluation, twelve cutting-edge approaches are chosen. The proposed method is also analyzed on underwater and low-light images. Extensive experiments indicate that the proposed method outperforms state-of-the-art methods of qualitative and quantitative performances.

Hazy Image Enhancement Using DCP and AHE Algorithms with YIQ Colour Space

Hazy Image Enhancement Using DCP and AHE Algorithms with YIQ Colour Space

Unsupervised Image Dedusting via a Cycle-Consistent Generative Adversarial Network

In sand–dust weather, the quality of the image is seriously degraded, which affects the ability of advanced applications to image using remote sensing. To improve the image quality and enhance the performance of image dedusting, we propose an end-to-end cyclic generative adversarial network (D-CycleGAN) for image dedusting, which does not require pairs of sand–dust images and corresponding ground truth images for training. In other words, we train the network in an unpaired way. Specifically, we designed a jointly optimized guided module (JOGM), comprised of the sandy guided synthesis module (SGSM) and the clean guided synthesis module (CGSM), which aim to jointly guide the generator through corresponding discriminator adversarials to reduce the color distortion and artifacts. JOGM can significantly improve image quality. We propose a network hidden layer adversarial branch to perform adversarials from inside the network, which better supervises the hidden layer to further improve the quality of the generated images. In addition, we improved the original CycleGAN loss function and propose a dual-scale semantic perception loss in feature space and a color identity-preserving loss in pixel space to constrain the network. Extensive experiments demonstrate that our proposed network model effectively removes sand dust, has better clarity and image quality, and outperforms the state-of-the-art techniques. In addition, the proposed method can help the target detection algorithm to improve its detection accuracy and capability, and our method generalizes well to the enhancement of underwater images and hazy images.

Physical-priors-guided DehazeFormer

Single-image dehazing is a challenging task in several machine-vision applications. Methods based on physical models and prior knowledge fail under certain conditions, resulting in defects such as color distortion. Transformer-based methods have a strong representation ability owing to their self-attention mechanism that can effectively obtain global information. However, this approach is computationally expensive, and its weak inductive bias capability increases the risk of overfitting on small-sample datasets. To address these problems, in this study, we propose a novel DehazeFormer guided by physical priors, named SwinTD-Net, which is trained according to supervised and self-supervised learning, and combines the advantages of physical priors and transformers. The proposed DehazeFormer learns features guided by physical priors, which improves the generalization ability of the network and enables it to achieve good restoration effects on both synthetic and real-world hazy images. In addition, we propose a more appropriate prior input to better use physical priors, and we design a multi-scale dark-light enhancement algorithm for image restoration post-processing, which can improve the visual perception quality for human observers while performing some local enhancements. Extensive experiments illustrate that the proposed method outperforms state-of-the-art methods. The code and pre-trained models are available to academics so that they can reproduce our results and test them (https://github.com/hocking-cloud/SwinTD_Net).

WMCP-EM: An integrated dehazing framework for visibility restoration in single image

Adopting local patches for varying haze conditions is crucial for optimizing the performance in single image dehazing. We propose a novel two-fold method with a prime focus on self-adaptive prior named Weighted Median Channel Prior (WMCP) to resolve the problems introduced by using a fixed size local-patch in the dehazing process. WMCP works by leveraging the spatially changing haze statistics such as inclusion–exclusion of related pixels for estimating depth-map in varying haze conditions. It is a scale-invariant technique that retains most of the information present in the local neighbourhood of the hazy input image for estimating scene depth, which traditional methods generally fail to preserve. In addition, an unsharp-masking based technique called edge-modulation (EM) promotes hidden or missing details such as microscopic edges and textures lost due to haze, making this scheme beneficial in ensuring a visually aesthetic and realistic dehazed image. This research also includes a set of ablation tests to assess the contributions of each module engaged in the dehazing process. We performed a comparative evaluation of our method with several state-of-the-art techniques, revealing its superiority in terms of visibility improvement and edge preservation, especially when the dense haze regions are taken into consideration.

An encoder‐decoder based CNN architecture using end to end dehaze and detection network for proper image visualization and detection

AbstractIndustrial sectors are reinventing in automation, stability, and robustness due to the rapid development of artificial intelligence technologies, resulting in significant increases in quality and production. Visual‐based sensor networks capture various views of the surrounding environment and are used to monitor industrial and transportation sectors. However, due to unclean suspended air particles that damage the whole monitoring and transportation systems, the visual quality of the images is degraded under adverse weather conditions. This research proposed a convolutional neural network‐based image dehazing and detection approach, called end to end dehaze and detection network (EDD‐N), for proper image visualization and detection. This network is trained on real‐time hazy images that are directly used to recover dehaze images without a transmission map. EDD‐N is robust, and accuracy is higher than any other proposed model. Finally, we conducted extensive experiments using real‐time foggy images. The quantitative and qualitative evaluations of the hazy dataset verify the proposed method's superiority over other dehazing methods. Moreover, the proposed method validated real‐time object detection tasks in adverse weather conditions and improved the intelligent transportation system.

OptiMD-3D DCNN: A Framework for Restoring the Haze-Free Images by Image Dehazing Techniques Using Heuristic Approach of Adaptive Lifting Wavelet Transform

The recently implemented illustrative techniques are focused to restore haze-free images by using learning approaches and also physical models. Managing finer details of the images when completely removing the haze is a complicated task especially in performing the hazing on single images. Generally, hazy images comprise distorted colors, poor contrast, and lower visibility, which lead to lower efficiency in the recognition, detection of objects, and classification of images. The main aspect of the proposed work is to restore haze-free images without any degradation in image quality. Initially, the required raw hazy images are acquired from the benchmark data sources. Further, these images undergo the decomposing stage, which is accomplished by adopting the Adaptive Lifting Wavelet Transform (ALWT), where the wavelet levels are optimized by using the improved hybrid algorithm of the Modified Tunicate Galactic Swarm Optimization (MTGSO). Once the decomposed images are obtained, Optimal Multiscale Dilation assisted 3D Deep Convolutional Neural Network (OptiMD-3D DCNN) is developed for the image dehazing task, in which the hyperparameters are optimally tuned using a hybrid MTGSO algorithm. At last, the effectiveness is examined and guaranteed that the recommended scheme shows higher performance in estimating with the classical methods.

Carbonaceous aerosols remote sensing from geostationary satellite observation, Part I: Algorithm development using critical reflectance

Current aerosol remote sensing without using multi-view satellite sensors still fail to retrieve its components directly, such as black carbon (BC) and organic carbon (OC) simultaneously due to the limitations on available observations. In this study, A new general carbonaceous aerosol retrieval strategy for geostationary single-view satellite observations is implemented based on a method of critical reflectance, to retrieve the parameters of BC and extra OC concentration from measured radiance without prior quantification of Aerosol Optical Thickness (AOT). The method is based on measuring the change in upward reflectance between clear and hazy days over a varying surface reflectance, with the aim of finding a balance between brightening due to scattering and darkening due to absorption. Since aerosols in the proposed algorithm are considered as a mixture of BC, OC and clustered shell aerosols, this balance can be used to retrieve concentration parameters of these components via radiative transfer models. Sensitivity tests indicate that different scenarios of BC and OC volume fractions can introduce certain sensitivity pattern on critical reflectance. The retrieval is achieved by minimizing the differences between measured critical reflectance and a set of look-up tables (LUTs). Random sampling consensus (RANSAC) is also used in the strategy to reduce the influence of clouds and anomalous pixels on the retrievals. Since critical reflectance should be calculated by comparing geometrically consistent clear and hazy images, the algorithm is more suitable for geostationary satellites with constant viewing geometries. An initial validation and application applied to Himawari images over the North China Plain (NCP) shows that the carbonaceous aerosol retrievals are highly consistent with the expectations in terms of spatial and temporal patterns. The retrievals of BC and OC concentration follow the daily fluctuations of the Aethalometers observations and feature small differences in mean values throughout the month. Additionally, the AOTs at 0.55 μm can be also satisfactorily reproduced based on the carbonaceous aerosol retrievals, resulting in a mean absolute error of 0.089 and a correlation coefficient of 0.870. The main errors in the method arise from shell model assumptions, RANSAC fitting bias, inconsistent clear reference AOT in LUT, and geometric inconsistency between the clear reference and hazy images. This work indicates that the BC and OC concentration in high resolution can be acquired through the geostationary single-view remote sensing for the further air quality and climate studies.

An end-to-end deep learning approach for real-time single image dehazing

Image dehazing methods can restore clean images from hazy images and are popularly used as a preprocessing step to improve performance in various image analysis tasks. In recent times, deep learning-based methods have been used to sharply increase the visual quality of restored images, but they require a long computation time. The processing time of image-dehazing methods is one of the important factors to be considered in order not to affect the latency of the main image analysis tasks such as detection and segmentation. We propose an end-to-end network model for real-time image dehazing. We devised a zoomed convolution group that processes computation-intensive operations with low resolution to decrease the processing time of the network model without performance degradation. Additionally, the zoomed convolution group adopts an efficient channel attention module to improve the performance of the network model. Thus, we designed a network model using a zoomed convolution group to progressively recover haze-free images using a coarse-to-fine strategy. By adjusting the sampling ratio and the number of convolution blocks that make up the convolution group, we distributed small and large computational complexities respectively in the early and later operational stages. The experimental results with the proposed method on a public dataset showed a real-time performance comparable to that of another state-of-the-art (SOTA) method. The proposed network’s peak-signal-to-noise ratio was 0.8 dB lower than that of the SOTA method, but the processing speed was 10.4 times faster.

Adversarial attack method against image classification based on haze perturbation

The adversarial attack is a cutting-edge technique used to study the vulnerability of deep neural networks (DNNs). However, most existing studies focus on the additive perturbation-based attack, which cannot represent real-world corruption and limit their applications. In particular, haze is a common natural phenomenon that significantly corrupts an image, which inevitably poses a potential threat to deep models. In this work, for the first attempt, we study the effects of haze on DNNs from the perspective of adversarial attacks and propose two adversarial haze attack methods. We first propose the optimization-based adversarial haze attack (OAdvHaze) that optimizes the parameters of the atmospheric scattering model with the guidance of a DNN to synthesize a hazy image, which leads to a high attack success rate. To achieve a more efficient attack, we further propose a predictive adversarial haze attack (PAdvHaze) employing a DNN to predict the hazing parameters through a one-step way. To validate the effectiveness of both methods, we conducted extensive experiments on two publicly available datasets, i.e., ILSVRC 2012 and NIPS 2017. OAdvHaze and PAdvHaze achieve comparable attack success rates and transferability to state-of-the-art attacks. This work would contribute to the evaluation and enhancement of the robustness of DNNs against haze perturbation that may happen in the real world.

Single Image Dehazing based on Additive Wavelet Transform

In this work, a single image dehazing method, which uses the multiscale products of additive wavelet transform as a prior is presented. In this method, first, the additive wavelet transform is applied to the hazy image to obtain its approximation and wavelet layers. Then the multiscale products of the approximation and detail layers of the input hazy image is calculated. The multiscale products of approximation and wavelet layers are summed up to obtain the proposed prior. Observations demonstrate that the proposed prior calculation keeps the detail information of the image, while detecting the haze. Using the proposed prior and commonly used hazy image model together, an efficient dehazing method is constructed. The comparisons between the proposed method and commonly used dehazing methods show that the proposed method has a better dehazing perfomrance than the traditional methods.

Enhancement of Marine Lantern’s Visibility under High Haze Using AI Camera and Sensor-Based Control System

This thesis describes research to prevent maritime safety accidents by notifying navigational signs when sea fog and haze occur in the marine environment. Artificial intelligence, a camera sensor, an embedded board, and an LED marine lantern were used to conduct the research. A deep learning-based dehaze model was learned by collecting real marine environment and open haze image data sets. By applying this learned model to the original hazy images, we obtained clear dehaze images. Comparing those two images, the concentration level of sea fog was derived into the PSNR and SSIM values. The brightness of the marine lantern was controlled through serial communication with the derived PSNR and SSIM values in a realized sea fog environment. As a result, it was possible to autonomously control the brightness of the marine lantern according to the concentration of sea fog, unlike the current marine lanterns, which adjust their brightness manually. This novel-developed lantern can efficiently utilize power consumption while enhancing its visibility. This method can be used for other fog concentration estimation systems at the embedded board level, so that applicable for local weather expectations, UAM navigation, and autonomous driving for marine ships.