Single Image Dehazing

Single Image Dehazing

We want to tackle the image/video pipeline across the board. Starting from the image capturing step, we focus on restoration applications that aim to recover the lost visual information due to poor acquisition conditions. Our work centers around the task of Single Image Dehazing which deals with recovering the scene information from images that have been affected by atmospheric haze.

Haze is an atmospheric phenomenon that arises when dust, smoke and other atmospheric particles absorb and scatter light. The radiance of objects attenuates as it traverses from the scene point to the camera and blends with additive atmospheric light, corrupting the visibility of the scene.

In computer vision, a hazy image, I(x), is determined by the amount of transmitted light that remains from the the clear scene, J(x), after traveling a distance, d(x), and the atmospheric light intensity, A(x).

I(x) = J(x) + J(x)t(x) + A(x)(1 – t(x))

The amount of particulates in the atmosphere cause the transmitted light to decay exponentially with the distance traveled. It is represented by the extinction coefficient, β.

t(x) = exp(-βd(x))

One can attempt to recover the clear image by solving the atmospheric light scattering model.

Here at DeepCamera, we dedicate our research for understanding, analyzing and improving the task of single image dehazing. We focus on building the foundations to foster algorithmic innovation in dehazing for both the classical image processing approach and the modern deep-learning domain.


Our team has created its very own repository of hazy datasets. Using the atmospheric scattering model, we were able to render 45 different levels of atmospheric haze on the Microsoft Common Objects in Context (COCO) dataset. Our collection holds over 150K fully annotated images, ready for benchmark tasks such as classification and object detection under different hazy conditions.

The single image dehazing literature has accumulated a wealth of algorithms over the years, and with them several evaluation metrics have spawned. We realize that while these evaluation metrics are successful in their intended purpose, they each address different aspects of the task. Our team has eliminated the need to trade off in evaluation by formulating a single objective performance indicator out the combination of these metrics.


Paper

2021

S. Panayi and A. Artusi

Hazing or Dehazing: the big dilemma for object detection

One of the biggest adversaries to the computer vision pipeline is bad weather which can deteriorate the visual quality of the captured images and lead to a decreased performance of tasks. Examples of such tasks can include image classification, object detection and semantic segmentation. To ameliorate this acquisition bottleneck, vision experts have developed restoration approaches which aim to recover the lost visual information due to the presence of poor climactic conditions such as atmospheric haze. The technique of single image dehazing has achieved great strides in producing aesthetically pleasing restorations to the human perception. However, it is important to establish whether these approaches bring the same merits to high-level vision tasks. To this end, we formulate a study around the task of object detection that aims to uncover the underlying relationship between this high-level task and atmospheric haze as well as examine the ability of the current dehazing process to enhance the detection performance. From our experiments we find that while there is a clear negative relationship between hazy conditions and detection performance, there is little help from the dehazing process to achieve the desired haze-free results.

Close

@inproceedings{panayi2021hazing,
title={Hazing or Dehazing: the big dilemma for object detection},
author={Panayi, Simoni and Artusi, Alessandro},
booktitle={2021 IEEE 23rd International Workshop on Multimedia Signal Processing (MMSP)},
pages={1–9},
year={2021},
organization={IEEE}
}

Close

2021

A. Artusi and K. A. Raftopoulos

A Framework for Objective Evaluation of Single Image De-hazing Techniques

Real-world environment, where images are acquired with digital camera, may be subject to sever climatic conditions such as haze that may drastically reduce the quality performance of sophisticated computer vision algorithms used for various tasks, e.g., tracking, detection, classification etc. Even though several single image de-hazing techniques have been recently proposed with many deep-learning approaches among them, a general statistical framework that would permit an objective performance evaluation has not been independently introduced yet. In this manuscript, certain performance metrics that emphasize different aspects of image quality, output ranges and polarity, are dentified and combined into a single performance indicator derived in an unbiased manner. A general methodology is thus introduced, as a framework for objective performance evaluation of current and future dehazing tasks, through an extensive comparison of 15 single image de-hazing techniques over a vast range of image data sets. The proposed unified framework shows several advantages in evaluating diverse and perceptually meaningful image features but also in elucidating future directions for improvement in image dehazing tasks.

Close

@ARTICLE{9437207,
author={Artusi, Alessandro and Raftopoulos, Konstantinos A.},
journal={IEEE Access},
title={A Framework for Objective Evaluation of Single Image De-Hazing Techniques},
year={2021},
volume={9},
number={},
pages={76564-76575},
doi={10.1109/ACCESS.2021.3082207}}

Close