Multi-scale depth information fusion network for image dehazing

Fan, Guodong; Hua, Zhen; Li, Jinjiang

doi:10.1007/s10489-021-02236-2

Cited by 24 publications

(7 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Fan et al [132] incorporated depth information in their multi-scale network (MSDFN). Specifically, they relied on the U-Net architecture.…”

Section: Multi-scale Based Dehazing Methodsmentioning

confidence: 99%

“…Method Short Description Year Dehazenet [118] 3-layer CNN, BReLU activation function AOD-Net [119] lightweight, transformed ASM Light-DehazeNet [120] lightweight, transformed ASM, CVR module FFA-Net [121] attention-based feature fusion structure CNNbased AECR-Net [122] AE-like, contrastive learning, feature fusion MSFFA-Net [123] multi-scale grid network, feature fusion GDNet [124] 3 sub-processes, multi-scale grid network MSCNN [125] 2 nets: coarse-and fine-scale MSCNN-HE [126] 3 nets: coarse-, fine-scale and holistic edge guided EMRA-Net [127] 2 nets: TRA-CNN and EA-CNN MSBDN [128] dense feature fusion module, boosted decoder FAMED-Net [129] 3 encoders at different scales, fusion module PGC [130] PGC and DRB blocks MSRA-Net [131] CIELAB, 2 subnets (luminance, chrominance) MSDFN [132] depth-aware dehazing DMPHN [133] non-homogeneous haze, multi-patch architecture TDN [134] 3 subnets: coarse-, fine-scale and haze density…”

Section: Categorymentioning

confidence: 99%

See 1 more Smart Citation

A Survey of Deep Learning-Based Image Restoration Methods for Enhancing Situational Awareness at Disaster Sites: The Cases of Rain, Snow and Haze

Karavarsamis

Gkika

Gkitsas

et al. 2022

Sensors

View full text Add to dashboard Cite

This survey article is concerned with the emergence of vision augmentation AI tools for enhancing the situational awareness of first responders (FRs) in rescue operations. More specifically, the article surveys three families of image restoration methods serving the purpose of vision augmentation under adverse weather conditions. These image restoration methods are: (a) deraining; (b) desnowing; (c) dehazing ones. The contribution of this article is a survey of the recent literature on these three problem families, focusing on the utilization of deep learning (DL) models and meeting the requirements of their application in rescue operations. A faceted taxonomy is introduced in past and recent literature including various DL architectures, loss functions and datasets. Although there are multiple surveys on recovering images degraded by natural phenomena, the literature lacks a comprehensive survey focused explicitly on assisting FRs. This paper aims to fill this gap by presenting existing methods in the literature, assessing their suitability for FR applications, and providing insights for future research directions.

show abstract

“…Fan et al [132] incorporated depth information in their multi-scale network (MSDFN). Specifically, they relied on the U-Net architecture.…”

Section: Multi-scale Based Dehazing Methodsmentioning

confidence: 99%

Section: Categorymentioning

confidence: 99%

A Survey of Deep Learning-Based Image Restoration Methods for Enhancing Situational Awareness at Disaster Sites: The Cases of Rain, Snow and Haze

Karavarsamis

Gkika

Gkitsas

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

“…In addition to the above neural network methods that learn features from the image space alone, some scholars have tried to add depth information to the network as prior knowledge to guide the network to accurately estimate the haze distribution. Fan et al [21] proposed a Multi-Scale Depth information Fusion Network (MSDFN). This method estimates the depth information of the haze image and fuses it with the haze image information at different scales for encoding and decoding to obtain a clear image.…”

Section: Deep Learning-based Dehazingmentioning

confidence: 99%

A multimodal feature fusion image dehazing method with scene depth prior

Zhang

Yan

Shuai

et al. 2023

IET Image Processing

View full text Add to dashboard Cite

Current dehazing networks usually only learn haze features in a single‐image colour space and often suffer from uneven dehazing, colour, and edge degradation when confronted with different scales of ground objects in the depth space of the scene. The authors propose a multimodal feature fusion image dehazing method with scene depth prior based on a decoder–encoder backbone network. The multimodal feature fusion module was first designed. In this module, affine transformation and polarized self‐attention mechanism are used to realize the fusion of image colour and depth prior feature, to improve the representation ability of the model for different scale ground haze feature in‐depth space. Then, the feature enhancement module (FEM) is added, and deformable convolution and difference convolution methods are used to enhance the representation ability of the model for the geometric and texture feature of the ground objects. The publicly available dehazing datasets are used for comparison and ablation experiments. The results show that compared with the existing classical dehazing networks, the peak signal‐to‐noise ratio (PSNR) and SSIM of the authors’ proposed method have been significantly improved, have a more uniform dehazing effect in different depth spaces, and maintain the colour and edge details of the ground objects very well.

show abstract

“…The additional information is mainly used to establish and solve the corresponding partial differential equation. The dehazing method based on the depth information uses the depth information of the foggy image for dehazing operation [7] . The most typical dehazing method based on prior information is the dark primary color prior algorithm proposed by He et al [8] .…”

Section: Introductionmentioning

confidence: 99%

Image haze removal algorithm based on nonsubsampled contourlet transform

Wang

Zhou

et al. 2023

AOPC 2022: Optical Sensing, Imaging, and Display Technology

View full text Add to dashboard Cite

Although the traditional dehazing algorithm can improve the clarity of hazy weather images, it may lead to the loss of many details and the distortion of color saturation in the process of processing. In order to overcome this defect and enhance the details of the image, a single image dehazing algorithm based on non-subsampling contourlet transform was proposed. First, the captured fog images are mapped from RGB space to HSI space, and the luminance channel map and saturation channel map are processed separately. The luminance channel image is decomposed by non-subsampling contourlet. The obtained high-frequency components are filtered by a guided filter, which can smooth the image while maintaining the edge. The obtained low-frequency components are processed by the single-scale Retinex algorithm to enhance the details of hazy areas in the image. A new luminance channel image is obtained through certain fusion rules and the inverse transformation. Then, the degradation model of the saturation component map is established. The parameters are estimated using the dark channel prior principle, and the estimated saturation map is obtained. Finally, the new luminance channel image, the estimated saturation image and the original hue channel image are inversely transformed to RGB space, resulting in a dehazed image. Experiments show that the method in this paper can solve the problem of color distortion in bright areas to a certain extent and the color saturation of the image, while keeping the overall outline structure of the image clearly and the edge details prominent. The visibility of the whole image is also improved, which is prior to the traditional detection algorithms.

show abstract

Multi-scale depth information fusion network for image dehazing

Cited by 24 publications

References 37 publications

A Survey of Deep Learning-Based Image Restoration Methods for Enhancing Situational Awareness at Disaster Sites: The Cases of Rain, Snow and Haze

A Survey of Deep Learning-Based Image Restoration Methods for Enhancing Situational Awareness at Disaster Sites: The Cases of Rain, Snow and Haze

A multimodal feature fusion image dehazing method with scene depth prior

Image haze removal algorithm based on nonsubsampled contourlet transform

Contact Info

Product

Resources

About