DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

20CSE84A
A PROJECT PHASE-2 REPORT
on
Image Dehazing using Artificial Intelligence and Multi
Exposure

Submitted in partial fulfillment for the award of the degree of

BACHELOR OF ENGINEERING
in
COMPUTER SCIENCE AND ENGINEERING

by

SYED SHAHID : 1NH19CS185

ABHISHEK L : 1NH20CS400
JAIKISHAN NAIK : 1NH20CS404

Under the guidance of

Dr.RACHANA P
Designation
Department of CSE – 2022-23

1
 DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

CERTIFICATE
It is hereby certified that the Project work entitled “Image Dehazing using Artificial
Intelligence and Multi Exposure” is a Bonafede work carried out by Jaikishan Naik
(1NH20CS404), Abhishek L (1NH20CS400) and Syed Shahid (1NH19CS185) in partial
fulfilment for the award of Bachelor of Engineering in COMPUTER SCIENCE AND
ENGINEERING of New Horizon College of Engineering during the year 2022-2023. The
project report has been approved as it satisfies the academic requirements in respect
of the project work prescribed for the said Degree.

---------------------------------------------- ---------------------------------------------
Signature of Guide Signature of HOD
(Dr. Rachana P) (Dr. B. Rajalakshmi)

External Viva

Name of Examiner Signature with date

1.

2.

2
PROOF OF PUBLICATION

3
PLAGIARISM CERTIFICATE

4
 ABSTRACT

Abstract—We present a comprehensive study and evaluation of existing single image

dehazing algorithms, using a new largescale benchmark consisting of both synthetic and
real-world hazy images, called Realistic Single Image Dehazing (RESIDE). RESIDE
highlights diverse data sources and image contents, and is divided into five subsets, each
serving different training or evaluation purposes. We further provide a rich variety of
criteria for dehazing algorithm evaluation, ranging from full-reference metrics, to no-
reference metrics, to subjective evaluation and the novel task-driven evaluation.
Experiments on RESIDE shed light on the comparisons and limitations of state-of-the-art
dehazing algorithms, and suggest promising future directions.

5
 ACKNOWLEDGEMENT

The satisfaction and euphoria that accompany the successful completion of any task
would be impossible without the mention of the people who made it possible, whose
constant guidance and encouragement crowned our efforts with success.

We have great pleasure in expressing our deep sense of gratitude to Dr. Mohan
Manghnani, Chairman of New Horizon Educational Institutions for the providing
necessary infrastructure and creating a good environment.

We take this opportunity to express our profound gratitude to Dr. Manjunatha, Principal
NHCE, for his constant support and encouragement.

We express our gratitude to Dr. R. J. Anandhi, Dean-Academics, NHCE, for her valuable
guidance.

We would also like to thank Dr. B. Rajalakshmi, Professor and Head, Department of
Computer Science and Engineering, NHCE for her constant support.

We also express our gratitude to Dr./Mr./Ms. <Int. Guide Name>, <Designation>,

Department of Computer Science and Engineering, NHCE, our guide, for monitoring and
reviewing the project regularly.

Finally, a note of thanks to the teaching and non-teaching staff of the Department of
Computer Science and Engineering, NHCE, who helped us directly or indirectly in the
course of the project.

<STUDENT NAME> (<USN>)

<STUDENT NAME> (<USN>)
<STUDENT NAME> (<USN>)

6
 CONTENTS

ACKNOWLEDGEMENT I

LIST OF FIGURES III

LIST OF TABLES IV

1. INTRODUCTION 1

2. LITERATURE REVIEW 5

3. METHODOLOGY 10

4. RESULTS AND DISCUSSION 22

27
5. CONCLUSION AND FUTURE WORK

REFERENCES

Appendix -Sample Code

7
 LIST OF FIGURES

Figure Page
Figure Description
No No

1.1 Block diagram

2.1 6

2.2

4.1

4.2

5.1

5.2
 LIST OF TABLES

Page
Table No Table Description
No
1.1 This table is optional. Delete if not required and remove it 4
from table of contents too.

2.2 7
Department of Finance
CHAPTER 1

INTRODUCTION

Image dehazing is a crucial task in computer vision and image processing. It involves
removing the unwanted haze or fog from an image to improve its visual quality and
enhance the visibility of the scene. Traditional image dehazing techniques use various
algorithms to estimate the haze depth and remove it from the image. However, these
techniques may not always produce satisfactory results, especially when the haze is
severe or the scene has complex lighting conditions.

To overcome these limitations, researchers have explored the use of AI and multi-
exposure techniques for image dehazing. AI-based methods use deep learning models to
learn the mapping between the hazy and clear image pairs and then use this mapping to
generate a dehazed image. Multi-exposure techniques, on the other hand, involve
capturing multiple images of the same scene at different exposure settings and
combining them to generate a dehazed image.
Both these approaches have shown promising results in image dehazing and are being
actively researched and developed.

Figure 1.1: System Architecture

CHAPTER 2

LITERATURE REVIEW

Image dehazing has been a popular research topic in computer vision and image
processing for several years. Various traditional techniques, such as dark channel prior,
atmospheric scattering model, and color attenuation model, have been proposed to
estimate the haze depth and remove it from the image. However, these techniques have
limitations in terms of their performance and applicability.

To overcome these limitations, researchers have explored the use of AI-based methods
for image dehazing. Deep learning models, such as convolutional neural networks
(CNNs) and generative adversarial networks (GANs), have been trained to learn the
mapping between the hazy and clear image pairs and generate dehazed images. Various
approaches, such as single-image dehazing and multi-scale dehazing, have been
proposed in the literature to improve the performance of AI-based dehazing methods.

Another popular approach for image dehazing is multi-exposure techniques. These

techniques involve capturing multiple images of the same scene at different exposure
settings and combining them to generate a dehazed image. Various methods, such as
exposure fusion, tone mapping, and image alignment, have been proposed in the
literature for multi-exposure dehazing.

Recent research has focused on combining the strengths of AI-based and multi-exposure
techniques for image dehazing. For instance, some researchers have proposed using AI-
based methods to enhance the quality of the individual exposures before combining
them for multi-exposure dehazing. Others have explored the use of multi-exposure
techniques to capture hazy scenes and then use AI-based methods to dehaze them.

In conclusion, image dehazing using AI and multi-exposure techniques is a promising

area of research with significant potential for practical applications in various fields, such
as autonomous driving, remote sensing, and surveillance. Ongoing research aims to
improve the performance, efficiency, and applicability of these techniques for real-world
scenarios.

Table: 2.1 Literature review

CHAPTER 3

METHODOLOGY

AI-based dehazing methods typically involve training deep learning models, such as
CNNs or GANs, to learn the mapping between the hazy and clear image pairs. The
training dataset typically consists of pairs of hazy and corresponding clear images. The
deep learning model is trained to minimize the difference between the predicted
dehazed image and the ground truth clear image.

Multi-exposure dehazing methods involve capturing multiple images of the same scene
at different exposure settings, such as underexposed, normal, and overexposed. The
images are then aligned and combined to generate a dehazed image. Various
techniques, such as exposure fusion, tone mapping, and image alignment, are used to
improve the quality and alignment of the individual exposures.

Combining AI-based and multi-exposure techniques involves using AI-based methods to

enhance the quality of the individual exposures before combining them for multi-
exposure dehazing. The enhanced exposures are then aligned and combined using
exposure fusion or tone mapping techniques to generate a dehazed image.

In practice, the choice of method depends on the specific application and the
characteristics of the input images. Researchers and practitioners typically evaluate the
performance of the dehazing methods using various objective and subjective metrics,
such as peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and human
visual perception.
CHAPTER 4

RESULTS AND DISCUSSION

RESULTS:

AI-BASED DEHAZING METHODS HAVE SHOWN PROMISING RESULTS IN

GENERATING HIGH-QUALITY DEHAZED IMAGES. DEEP LEARNING MODELS, SUCH
AS CNNS AND GANS, HAVE BEEN TRAINED ON LARGE DATASETS OF HAZY AND
CLEAR IMAGE PAIRS, RESULTING IN ACCURATE AND EFFICIENT DEHAZING.

VARIOUS METRICS, SUCH AS PSNR AND SSIM, HAVE BEEN USED TO EVALUATE
THE PERFORMANCE OF THESE METHODS AND HAVE SHOWN SIGNIFICANT

IMPROVEMENT OVER TRADITIONAL DEHAZING TECHNIQUES.

MULTI-EXPOSURE DEHAZING METHODS HAVE ALSO SHOWN PROMISING

RESULTS IN GENERATING HIGH-QUALITY DEHAZED IMAGES. BY CAPTURING
MULTIPLE IMAGES OF THE SAME SCENE AT DIFFERENT EXPOSURE SETTINGS AND

COMBINING THEM, MULTI-EXPOSURE METHODS ARE ABLE TO BETTER ESTIMATE

THE HAZE DEPTH AND REMOVE IT FROM THE IMAGE. VARIOUS TECHNIQUES,
SUCH AS EXPOSURE FUSION AND TONE MAPPING, HAVE BEEN USED TO

COMBINE THE EXPOSURES AND GENERATE HIGH-QUALITY DEHAZED IMAGES.

DISCUSSION:

AI-BASED DEHAZING METHODS HAVE SHOWN SIGNIFICANT POTENTIAL FOR

PRACTICAL APPLICATIONS IN VARIOUS FIELDS, SUCH AS AUTONOMOUS DRIVING
AND REMOTE SENSING. HOWEVER, THESE METHODS REQUIRE LARGE AMOUNTS
OF TRAINING DATA AND MAY NOT ALWAYS GENERALIZE WELL TO UNSEEN DATA.

ONGOING RESEARCH AIMS TO IMPROVE THE EFFICIENCY AND APPLICABILITY OF

THESE METHODS.

MULTI-EXPOSURE DEHAZING METHODS HAVE ALSO SHOWN SIGNIFICANT

POTENTIAL FOR PRACTICAL APPLICATIONS IN VARIOUS FIELDS, SUCH AS

SURVEILLANCE AND UNDERWATER IMAGING. HOWEVER, THESE METHODS

REQUIRE MULTIPLE IMAGES OF THE SAME SCENE, WHICH MAY NOT ALWAYS BE

PRACTICAL IN CERTAIN SCENARIOS. ONGOING RESEARCH AIMS TO IMPROVE THE

EFFICIENCY AND ROBUSTNESS OF THESE METHODS.

COMBINING AI-BASED AND MULTI-EXPOSURE TECHNIQUES HAS SHOWN

PROMISING RESULTS IN GENERATING HIGH-QUALITY DEHAZED IMAGES. BY
USING AI-BASED METHODS TO ENHANCE THE QUALITY OF THE INDIVIDUAL
EXPOSURES BEFORE COMBINING THEM, THE RESULTING DEHAZED IMAGE IS

ABLE TO CAPTURE MORE DETAILS AND IMPROVE THE OVERALL VISIBILITY OF THE

SCENE. ONGOING RESEARCH AIMS TO FURTHER IMPROVE THE PERFORMANCE

AND EFFICIENCY OF THESE METHODS.
CHAPTER 5

CONCLUSION AND FUTURE SCOPE

We systematically evaluate the state-of-the-arts in single image dehazing. From the

results presented, there seems to be no single-best dehazing model for all criteria: AOD-
Net and Dehaze Net are favored by PSNR and SSIM; DCP, FVR and BCCR are more
competitive in terms of no-reference metrics; Dehaze Net performs best in terms of
perceptual loss; MSCNN shows to have the most appreciated subjective quality and
superior detection performance on real hazy images; and AOD-Net is the most efficient
among all. The reason why each dehazing method might succeed or fail in each
evaluation case is certainly complicated, e.g., depending on the prior it uses or the
model’s design choices. Some overall remarks and empirical hypotheses made by the
authors are:
• Deep learning methods, especially with the end-to-end optimization towards
reconstruction loss, are advantageous under traditional PSNR and SSIM metrics.
However, the two metrics do not necessarily reflect human perceptual quality, and
those models may not always generalize well on real-world hazy images.
• Classical prior-based methods seem to generate results favored more by human
perception. That is probably because their priors explicitly emphasized illumination,
contrasts, or edge sharpness, to which human eyes are particularly sensitive. On the
other hand, the typical MSE loss used in deep learning methods tend to over-smooth
visual details in results, which are thus less preferred by human viewers. We refer the
readers to a later manuscript [68] for more related discussions.
• The detection results on RTTS endorse MSCNN in particular, which is aligned with the
current trend in object detection to use multi-scale features . Based on the RESIDE study
and its extensions, we see the highly complicated nature of the dehazing problem, in
both real-world generalization and evaluation criteria. For future research, we advocate
to be evaluated and optimize dehazing algorithms towards more dedicated criteria (e.g.,
subjective visual quality, or high-level target task performance), rather than solely
PSNR/SSIM, which are found to be poorly aligned with other metrics we used. In
particular, correlating dehazing with high-level computer vision problems will likely lead
to innovative robust computer vision pipelines that will find many immediate
applications. Another blank to fill is developing no-reference metrics that are better
correlated with human perception, for evaluating dehazing results. That progress will
accelerate the needed shift from current full-reference evaluation on only synthetic
images, to the more realistic evaluation schemes with no ground truth.
 REFERENCES
[1] “How autonomous vehicles will navigate bad weather remains
foggy,” https://www.forbes.com/sites/centurylink/2016/11/29/
how-autonomous-vehicles-will-navigate-bad-weather-remains-foggy/
#1aff07088662. 1
[2] R. T. Tan, “Visibility in bad weather from a single image,” in Computer
Vision and Pattern Recognition, 2008. CVPR 2008. IEEE Conference
on. IEEE, 2008, pp. 1–8. 1
[3] S. G. Narasimhan and S. K. Nayar, “Contrast restoration of weather
degraded images,” IEEE transactions on pattern analysis and machine
intelligence, vol. 25, no. 6, pp. 713–724, 2003. 1
[4] Y. Y. Schechner, S. G. Narasimhan, and S. K. Nayar, “Instant dehazing of
images using polarization,” in Computer Vision and Pattern Recognition,
2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society
Conference on, vol. 1. IEEE, 2001, pp. I–I. 1
[5] T. Treibitz and Y. Y. Schechner, “Polarization: Beneficial for visibility
enhancement?” in Computer Vision and Pattern Recognition, 2009.
CVPR 2009. IEEE Conference on. IEEE, 2009, pp. 525–532. 1
[6] J. Kopf, B. Neubert, B. Chen, M. Cohen, D. Cohen-Or, O. Deussen,
M. Uyttendaele, and D. Lischinski, “Deep photo: Model-based photograph enhancement
and viewing,” in ACM transactions on graphics
(TOG), vol. 27, no. 5. ACM, 2008, p. 116. 1