-
Scientific Journal Impact Factor (SJIF): 1.711
International Journal of Modern Trends in Engineering
and Research www.ijmter.com
@IJMTER-2014, All rights Reserved 334
e-ISSN: 2349-9745
p-ISSN: 2393-8161
Survey on Haze Removal Techniques
Lipakshee Bisen, Prof. Mr. Amit Dravid
G.H.Raisoni institute of engineering and
technology(GHRIET),Pune
ABSTRACT : This paper analyzed different haze removal methods.
Haze causes trouble to
many computer graphics/vision applications as it reduces the
visibility of the scene. Air light and
attenuation are two basic phenomena of haze. air light enhances
the whiteness in scene and on
the other hand attenuation reduces the contrast. the colour and
contrast of the scene is recovered
by haze removal techniques. many applications like object
detection , surveillance, consumer
electronics etc. apply haze removal techniques. this paper
widely focuses on the methods of
effectively eliminating haze from digital images. it also
indicates the demerits of current
techniques. Keywords: Image Dehazing, ICA, Depth, DCP, Contrast
enhancement, Polarizers
I. INTRODUCTION
The bad weather conditions may demean the quality of the images
of outdoor scenes. It is an
annoying problem for a photographer who captures images but the
images results into change of
colours, blur image, etc. This is an ultimatum to reliability of
many applications. The unwanted
condition is caused by the atmospheric conditions like haze[1]
and fog, which blurs the captured
scene. Always the air is misted by some added particles which
are scattered around, and hence,
the reflected light is also scattered which results in less
visibility of distant objects. The scattering
is caused by two basic events namely attenuation and airlight
[2, 1]. This occurrence affects the
normal work of automatic monitoring system, outdoor recognition
system, tracking &
segmentation and intelligent transportation system. In the last
few years, a technique has gained popularity and this is known as
restoration of
images that are taken into bad atmospheric conditions. This
specific task has become important
for several outdoor applications such as remote sensing,
intelligent vehicles, object recognition
and surveillance. The processing of recorded bands of reflected
light is done in order to restore
the outputs in remote sensing systems. Generally, haze may
enervate the light reflected from the
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 335
scenes and in fact merge some additional light in the
environment. This effect of haze can be
reduced by haze removal technique by improving the reflected
light and avoiding the merging of
additional light in the atmosphere. There are several haze
removal techniques such as
polarization[3,4] , independent component analysis, dark channel
prior etc.
II. THEORETICAL RELEVANCE
Haze removal techniques are gaining popularity due to its
availability in many classifications.
These methods can be used to construct a high quality, noise
free, dehaze images. The
classifications are done in two major types image segmentation
and image restoration. Due to the
presence of fog, mist, haze into the atmosphere the images
captured of outdoor scenes may have
a low quality. In many surveillance and transportation area haze
removation is important task.
This approach includes the analysis of scene, extraction of
useful information and then detecting
the image. Mostly in a bad weather condition the light that is
visible is captivated and is scattered
by other particles or raindrops. This prototype is engaged in
many haze removal approaches and
is exhibited as,
I(x) = J(x) t(x) + A (1 t(x)) ----------------------------- (1)
Where, I is the haze image on the three R, G, B color channels. J
is the scene without haze, t is
the transmission coefficient to describe the percentage of light
that can penetrate through haze,
and A is the atmospheric light. Using this atmospheric
scattering model to recover the scene J,
the main challenge of haze removal is to estimate the
atmospheric light A and the transmission t
from the source image I properly. The dark channel prior is
based on the following observation.
On haze-free outdoor images in which most of the non-sky patches
contain at least one color
channel has very low intensity at some pixels. By using this it
requires some extensive and
complex computations, such as huge matrix multiplication or
division, sort, exponent, and
floating point operations. We further investigate some various
haze removal methods like
multiple image scheme, single image with depth image scheme and
single image scheme. A. Haze Removal methods Haze removal methods
can be used to construct a high quality, noise free, dehaze images.
The
classifications are done in two major types image segmentation
and image restoration. 1) Image Segmentation: As the name suggests,
image segmentation is the process of segregation of a digital image
into
multiple segments. The purpose of segmentation is to clarify
and/or change the representation of
an image into something that is more meaningful and easier to
analyze. This technique is
primarily used to locate objects and boundaries in images.
Actually image
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 336
segmentation is the process of assigning a label to every pixel
in an image such that pixels with
the same label share certain visual characteristics. 2) Image
Restoration: Image restoration is the process of taking a
corrupted/noisy image and evaluating the clean
original image. The image corruption is caused by many reasons
such as motion blurs, noise,
camera miss-focus image, etc. The process of image restoration
is very different from the
concept of image enhancement. In the image enhancement process,
the designing of the image is
done to highlight the feature of the captures image resulting
the image more pleasing to the
observer. From a scientific point of view there is no necessity
to produce realistic data. No
previous methods are used in image enhancement techniques that
are provided by Imaging
packages. In fact with this approach, noise can be removed
effectively by relinquishing some
image resolution. But this phenomenon is not always accepted by
many applications. As it is in
Fluorescence Microscpe seen the resolution in the z-direction is
not good. But the image
restoration techniques recover the haze image with better
quality and brightness. For recovering
the object, there must be more advanced image processing
techniques available. Increasing
resolution especially in the axial direction removes noise and
increasing contrast. B. Haze Removal using dark channel prior :- A
remarkable progress in single image haze removal technique is
observed in recent days. The
use of stronger assumptions or prior methods may lead to the
success of haze removal technique.
Different researchers can use different methods to remove haze
from the images. In [5], the
author has used a soft matting algorithm to remove the haze. But
this model is physically invalid
and the assumption of constant air light may be unsuitable when
the sunlight is very influential.
Tarel uses image restoration technique to recover the haze. The
author in [6], estimates the
albedo of the scene and the medium transmission under the
assumption that the transmission and
the surface shading are locally uncorrelated. This technique is
physically possible and can give
imposing results. But there are some drawbacks of this system,
as it cannot dark hazy images and
it may also fail when the assumption is broken.
III. DEHAZING METHODS Haze removal techniques can be classified
into two categories which are as follows : 1) multiple
image dehazing method 2) single image dehazing methods
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 337
2.1. Multiple Image Dehazing Methods This method prefers two or
more images or multiple images of same scene. It completely
avoids
unknown and attains known methods only. Explanation of the
methods under this category is
given below: 2.1.1 Weather condition based method This
techniques utilizes multiple images(7,2,8) adapted from various
weather circumstances. In
the basic method the variations of two or more images of same
scene are considered. These
images possess distinct characteristics of the contributing
medium on the one hand it enhances
visibility but on the other hand it also make the user wait till
the characteristics of the medium
changes. This techniques does not immediately deliver the
results. this methods is also unable to
handle dynamic scenes.
(a) Hazy Image (c) Dehazed Image
(b) Hazy Image (d) Clear Weather Image
Figure 1. Multiple Image dehazing
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 338
Haze Removal
Techniques
Multiple Image Haze Single Image Haze
removal technique removal technique
Weather condition Contrast maximization
based technique technique
Polarization based Independent
technique component analysis
Depth map based
technique Dark channel prior
technique
Antistrophic Diffusion
technique
Figure 2. Classification of Haze removal methods 2.1.2
Polarization based method This methods having different
polarization filters(9,10) but of the same scene are
considered.
First of all, in this method distinct images are captured by
ratting a polarizing filter. but the
treatment results of dynamic scene is not really good. The
demerits of this method are- It require special equipment like
polarizers.
It is not applicable to dynamic scene where changes are more
quick than filter rotation.
It does not furnish better results.
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 339
(a) Best Polarization State (b)Worst Polarization State Figure
3. Image dehazing
using polarizing filters 2.1.3 Depth map based method This
method depth information for haze removal is considered. here we
consider 3D geometrical
model(2, 7, 10) of scene is given by certain databases like
google maps and also considers the
texture of the scene is supplied (from aerial photos or
satellite pictures). This 3D model aligns
hazy image and provides the scene depth[11]. This method wants
interaction to align 3D model
[12] with the scene and also provide accurate results. In this
method special equipments are not
needed. The demerits of this method are:
This method require user interaction
This method is not automatic
This method needs an estimation of more parameters, and the
extra information not easy
to adopt.
(a) Hazy image (b) 3D structural model (c) Dehazed Image
Figure 4. Depth map based method
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 340
2.2 Single image dehazing method Unlike previous method this
method only want a single input image(1,13). This method
depends
upon statistical assumption [14] and essence of the scene and it
also reclaim the scene data based on
last data from single image. This method is now attracting many
researchers. Following are the
methods which come under this category. 2.2.1 Contrast
maximization method Haze reduces the contrast elimination of the
haze increase the contrast of the image. This method
increases the contrast under the constraint. As this method does
not physically enhance depth or
brightness, the resultant image have greater saturation values.
The results also constitute halo effects
at depth discontinuities.
a) Hazy Image (b) Restored Image
Figure 5. Contrast Maximization Method
2.2.2 Independent Component Analysis(ICA) ICA is a statistical
method of dividing two additional components from a single. this
method is used
by fatal [13] and it is based on the assumption that surface
shading are statistically uncorrelated in
local patch. this approach provides good results and physically
valid , but one of the most important
disadvantage of this method is that it does not give paper
result in case of dense haze.
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 341
(a) Hazy Image (b) Haze-free image
Figure 6. Independent component analysis
2.2.3 Dark Channel prior The dark channel prior [14] is based on
the statistics of outdoor haze-free images. In most of the non-
sky patches, at least one color channel (RGB) has very low
intensity at some pixels (called dark
pixels). These dark pixels provide the estimation of haze
transmission. This approach is physically
valid and work well in dense haze. When the scene objects are
similar to the air light then it is
invalid.
(a)Hazy Image (b) Recovered Depth map (c) Haze-free image
Figure 7. Dark channel prior
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 342
2.2.4 Anisotropic diffusion Anisotropic diffusion [15] is a
technique that reduces haze without removing image parts such
as
edges, lines or other details that are essential for the
understanding of the image. Its flexibility permits
to combine smoothing properties with image enhancement
qualities. Tripathi [16] present an
algorithm uses anisotropic diffusion for refining air light map
from dark channel prior. Antistrophic
diffusion is used to smooth the airlight map. It performs well
in case of heavy fog.
IV. RELATED WORK
The author Schechner and et al in his paper has given his work,
which is based on the fact that the
scattered airlight is partially polarized. This airlight is
scattered by the atmospheric particles. But
only the polarization filtering cannot remove the haze effect.
In the proposed work, the image
formation process is shown where the image is a clean image. The
polarization effect is considered
and the inverting process is utilized, where it outputs into a
haze free image. Two components are
used to compose the image, one is known as scene radiance and
the other is airlight. Scene radiance
is in the absence of haze. And airlight is the ambient light
that is scattered towards the viewer. For
recovering the two components, there is a need for two
non-dependent images. And these images can
easily be acquired because airlight is partially polarized. This
approach can be immediately applied.
It does not require the change in weather conditions. The images
that are taken by a polarizer uses the
concept of polarization filtering. This polarization filtering
is used in photography across haze. The
aim of polarization filtering is to improve the contrast of the
input image.
In [13] Fattal proposed a new approach for single image dehazing
which try to implement haze free
image from the hazy image. Fattal formulated the refined image
formation model that relates to the
surface shading and the transmission function.
He and et al [14] dark channel prior is based on prior
assumption. It has been observed that in most of
the local regions which do not cover the sky, some pixels have
very low intensity in at least one color
(RGB) channel and these pixels are known as the dark pixels. In
hazy images the intensity of the dark
pixels in that color channel is basically contributed by the
airlight and these dark pixels are used to
estimate the haze transmission. After estimation of the
transmission map for each pixel, combining
with the haze imaging model and soft matting technique [17] to
recover a high quality haze free
image.
Ancuti and et al. [18] is described haze is atmospheric term
which degrades the outdoor image
visibility under the bad weather condition. This paper describes
single image dehazing approach
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 343
which is based on fusion strategy and it has been derived from
the original hazy image inputs by
applying a white balance and contrast enhancing procedure. The
fusion enhancement technique
estimates perceptual based qualities known as the weight maps
for each pixel in the image. These
weight maps control the contribution of each input to the final
obtained result. Different weight maps
like luminance, chromaticity and saliency are computed and to
minimize the artifacts produced
during the weight maps, the multiscale approach uses the
laplacian pyramid representations
combination with gaussian pyramids of normalized weights. As
this approach tries to minimize the
artifacts per pixel based has a greater improvement rather than
considering a patch based method due
to the assumption of contrast airlight in the patch.
Xie and et al [19] paper describes the dehazing process using
dark channel prior and multi-scale
retinex. This paper also focuses on the approach which provides
the automatic and fast acquisition of
transmission map of the scene. The proposed approach is based on
the implementing the multi scale
retinex algorithm on the luminance component in YCbCr space
of the input image to get the pseudo transmission map. The
obtained pseudo transmission map is very
much similar to the transmission map obtained by using the dark
channel prior by He et.al[14].
Combining the haze imaging model and the dark channel prior, a
high quality haze free image is
recovered.The input hazy image has been transformed from RGB
color space to YCbCr space and
then by using the multiscale retinex algorithm, on the luminance
component of the transformed
image with some adjustment to get the transmission map. Then
combining both the haze image
model and the retinex algorithm a better haze free image is
recovered.
Schaul and et al. [20] focused on the fact that in outdoor
photography, the distant object are appeared
as blurred and loses its color and visibility due to the
degradation level affected by the atmospheric
haze. In this paper the key idea is used to fusion of the
visible and a near-infrared image of the given
input image to obtain a dehazed image and it also describes the
multiresolution approach using the
edge preserving filter to minimize the artifacts those are
produced during the dehazing process.
IV CONCLUSION
Many vision applications apply haze removal algorithms. In past
few days it was discovered that
researchers have neglected many problems like no technique is
appropriate for distinct circumstances.
We have came to the conclusion that the presented methods have
ignored the techniques to diminish
the noise problem which is given in the output images of the
current fog removal algorithms. The issue
of lack of uniformity and over illumination is also an problem
for
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 344
dehazing the methods. so it is essential to rectify the current
techniques in such a manner that rectified method will work
efficiently.
V REFERENCES
[1] Tan, Robby T, "Visibility in bad weather from a single image
IEEE Conference on
Computer Vision and Pattern Recognition, CVPR, pp. 1-8, Year
2008.
[2] Narasimhan, Srinivasa G. and Shree K. Nayar, "Chromatic
framework for vision in bad
weather", The Proceedings of IEEE Conference on Computer Vision
and Pattern Recognition,
vol. 1, pp. 598-
605, 2000.
[3] Y. Y. Schechner, S. G. Narasimhan, and S. K. Nayar. Instant
dehazing of images using
polarization. CVPR, 1:325, 2001.
[4] S. Shwartz, E. Namer, and Y. Y. Schechner. Blind haze
separation. CVPR, 2:19841991, 2006.
[5] K. He, J. Sun, and X. Tang, Guided image filtering, in Proc.
11th Eur. Conf. Comput. Vis.
Part I, pp. 1 14.,2010.
[6] J.-P. Tarel and N. Hautire, Fast visibility restoration from
a single color or gray level image,
in Proc. IEEE Int. Conf. Comput. Vision, pp. 22012208,Sep.
2009.
[7] Narasimhan, Srinivasa G. and Shree K. Nayar, "Contrast
restoration of weather degraded
images", IEEE Transactions on Pattern Analysis and Machine
Intelligence, vol. 25, no. 6,pp.
713- 724, 2003.
[8] Nayar, Shree K. and Srinivasa G. Narasimhan, "Vision in bad
weather", The Proceedings of the
Seventh IEEE International Conference on Computer Vision, vol.
2, pp. 820-827, 1999.
[9] Schechner, Yoav Y., Srinivasa G. Narasimhan and Shree K.
Nayar, "Instant dehazing of images
using polarization", The Proceedings of the IEEE Computer
Society Conference on Computer
Vision and Pattern Recognition ( CVPR), vol. 1, pp. I-325,
2001.
[10] Shwartz, Sarit, Einav Namer and Yoav Y. Schechner, "Blind
haze separation", IEEE Computer
Society Conference on Computer Vision and Pattern Recognition,
vol. 2, pp. 1984-1991, 2006.
[11] Hautire, Nicolas, J-P. Tarel and Didier Aubert, "Towards
fog-free in-vehicle vision systems
through contrast restoration", IEEE Conference on Computer
Vision and Pattern Recognition,
(CVPR), pp. 1-8, 2007.
-
International Journal of Modern Trends in Engineering and
Research (IJMTER) Volume 01, Issue 05, [November - 2014] e-ISSN:
2349-9745, p-ISSN: 2393-8161
@IJMTER-2014, All rights Reserved 345
[12] Kopf, Johannes, Boris Neubert, Billy Chen, Michael Cohen,
Daniel Cohen-Or, Oliver Deussen,
Matt Uyttendaele, and Dani Lischinski. "Deep photo: Model-based
photograph enhancement
and viewing." In ACM Transactions on Graphics (TOG), vol. 27,
no. 5, p. 116, 2008. [13] Fattal, Raanan. "Single image dehazing."
In ACM Transactions on Graphics (TOG), vol. 27,
no. 3, p. 72, 2008. [14] He, Kaiming, Jian Sun and Xiaoou Tang,
"Single image haze removal using dark channel
prior.", IEEE Transactions on Pattern Analysis and Machine
Intelligence, vol.33, no. 12, pp.
2341-2353, 2011. [15] Tripathi, and S. Mukhopadhyay, "Single
image fog removal using anisotropic diffusion.",
Image Processing, Vol. 6, no. 7, pp. 966-975,2012. [16] Nayar,
Shree K. and Srinivasa G. Narasimhan, "Vision in bad weather", The
Proceedings of the
Seventh IEEE International Conference on Computer Vision, vol.
2, pp. 820-827, 1999. [17] Levin, D. Lischinski, and Y. Weiss, A
closed Form Solution to Natural Image Matting,
Proc. IEEE Conf. Computer Vision and Pattern Recognition, vol.
1, pp. 61-68, 2006 [18] Levin, D. Lischinski, and Y. Weiss, A
closed Form Solution to Natural Image Matting,
Proc. IEEE Conf. Computer Vision and Pattern Recognition, vol.
1, pp. 61-68, 2006 [19] Xie, Bin, Fan Guo, and Zixing Cai.
"Improved single image dehazing using dark channel prior
and multi-scale Retinex."Intelligent System Design and
Engineering Application (ISDEA),
2010 International Conference Vol. 1. IEEE, 2010. [20] Schaul,
Lex, Clment Fredembach, and Sabine Susstrunk. "Color image dehazing
using the
near-infrared." International Conference on Image Processing
(ICIP), 2009 16th IEEE
International Conference on. IEEE, 2009.