Int. J. Electron. Commun.

(AE) 69 (2015) 17731782

Contents lists available at ScienceDirect

International Journal of Electronics and

Communications (AE)
journal homepage: www.elsevier.com/locate/aeue

Fast single image and video deweathering using look-up-table

approach
Apurva Kumari , Subhendu Kumar Sahoo
BITS-Pilani, Hyderabad Campus, Hyderabad 500078, India

a r t i c l e

i n f o

Article history:
Received 15 December 2014
Accepted 4 September 2015
Keywords:
Gamma correction
Look-up-table method
Bad weather condition

a b s t r a c t
The bad weather conditions including fog, haze, rain and snow make the acquired image contrast lower.
This makes computer vision application fail to detect or recognize objects. In this paper, a fast and
better technique for deweathering the image and video is presented. Here, we proposed an approach
of gamma correction using look-up-table methods to obtain higher contrast images. The experimental
results demonstrated that this method could effectively remove the bad weather condition and enhance
the contrast of the input images and videos. The use of look-up-table (LUT) in these images makes our
approach appreciably faster in comparison to the recent reported approach of Tarel and Wang, respectively. Moreover, we extend the static image to real-time video deweathering. The main advantage of the
proposed approach compared with others is its higher speed which allows the reliable real-time video
processing.
2015 Elsevier GmbH. All rights reserved.

1. Introduction
Imaging in poor weather is often severely degraded by scattering due to suspended particles in the atmosphere such as haze,
fog, mist, rain and snow. Poor visibility in bad weather is a major
problem for many applications such as intelligent vehicles, outdoor
object recognition and remote sensing systems, etc. The real time
deweathering is very important for many such similar applications.
In our proposed method, image enhancement technique is used for
improving the contrast of the image. This method does not analyze
the physical causes of image degradation caused by bad weather
conditions, but it is a process that aims to improve the quality of
the image according to the visual experience, which is belonging to
the areas of the image enhancement. This method is important for
enhancing the contrast of the images as well as speed.
Recently, a lot of efforts have been made to develop efcient
deweathering algorithms. Several dehazing algorithms have been
proposed to estimate object depths using multiple images or additional information. Object depths are estimated from two images,
which are captured in different weather conditions [1,2] or with
different degrees of polarization [3,4]. But this approach is unsuitable for real-time image deweathering as it processes two images
taken under different weather conditions. Single image dehazing
algorithms have been developed to overcome the limitation of

Corresponding author. Tel.: +91 4066303569.

E-mail address: [email protected] (A. Kumari).
http://dx.doi.org/10.1016/j.aeue.2015.09.001
1434-8411/ 2015 Elsevier GmbH. All rights reserved.

multiple image dehazing approaches. These algorithms make use

of strong assumptions to remove haze from a single image. Tan [5]
maximized the contrast of a hazy image, assuming that a haze-free
image has a higher contrast ratio than the hazy image. Tans algorithm, however, tends to overcompensate for the reduced contrast,
yielding halo artifacts. The above conclusion can be observed from
Fig. 1 (b), wherein there are halo artifacts present in the mountain.
Fattal [6] exploits the fact that the transmission and scene albedo
are locally uncorrelated to dehaze the image.
This approach is physically sound and can produce impressive
results. However, it is deeply based on the color and thus cannot
deal with a gray level image. He et al. [7] proposed a novel algorithm
based on dark channel prior (DCP) which is physically valid and is
able to handle distant objects in heavily hazy images. He found that
some pixels have very low intensity in at least one color (RGB) channel. His method shows the best quality of result with haze among
the image restoration methods (as shown in Fig. 1(e)), where the
soft matting algorithm to rene the transmission map was used.
This matting algorithm requires numerous data to obtain the exact
transmission at discontinuous edge of depth map. Therefore, it is
difcult for real time processing. To overcome these problems, Tarel
and Hautiere [8] proposed a novel algorithm based on median lter. The algorithm is controlled only by a few parameters such
as atmospheric veil inference, image restoration, smoothing and
tone mapping. This is based on the assumption that fog is white.
Though Tarels method is possible for real time processing, but
the halos cannot be completely eliminated which can be observed
from Fig. 1(h). Recently, He et al. [9] proposed a guided lter, and

1774

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

Fig. 1. (a) Original foggy images, (b) Tan result, (c) our result, (d) original image (e), He result, (f) our result, (g) original image, (h) Tarel result, (i) our result, (j) original image,
(k) GDCP, (l) our result.

found that the output of a guided lter could be approximate solution of the soft matting algorithm. This algorithm reduces the time
complexity. However, this may lead to image blurring because the
original foggy image was chosen as the reference image shown in
Fig. 1(k). Recently, Wang et al. [10] presents a new method for fast
single haze image enhancement without using any extra information. This approach simultaneously dehazes image and enhances
sharpness by means of individual treatment of the model component and the residual. This method requires an image enhancement
technique to remove the halo artifacts.
This work proposes the real time approach for enhancing the
contrast of the weather degraded images. The image enhancement
method appeals better results with less processing time. We will
discuss the proposed approach which is based on the enhancement method. This approach removes the haze and fog by simply
applying the gamma correction using look-up-table method. The
complexity of our approach is lower than most of the previous
strategies as the transmission map estimation is not required. The
proposed method is much faster than above discussed approaches
such as Tan [5], Fattal [6], He et al. [7], Tarel et al. [8], GDCP method
[9] and Wang et al. [10] as it is based on the look-up-table method
and improves the quality of the images. Proposed method, does not
require any pre/post processing steps.
The rest of the paper is organized as follows: Section 2 provides a
brief discussion of related techniques that deal with deweathering.
Section 3 presents our proposed method in detail for deweathering
the images. In Section 4, the effectiveness of the proposed method
is supported by comparing the experimental results obtained
through use of our method to those obtained via existing methods.
Finally, Section 5 concludes this work.
2. Previous works
Enhancing images represent a fundamental task in many image
processing and vision applications. As a particularly challenging
case, restoring the weather degraded images requires specic
strategies and therefore an important variety of methods have
emerged to solve this problem. Given the importance of dehazing

algorithms, many studies on dehazing have been conducted. Previous dehazing research can be divided into two categories: image
enhancement methods and image restoration methods. Although
the image restoration based algorithms can achieve better results,
but these methods are usually complex. As a result, the algorithms will cost much more running time and it is hard to apply
to the real time systems. In contrast with the image restoration
based algorithm, the image processing based algorithms simply
process the image and enhance the contrast. More importantly,
the running time, cost much less than the image restoration based
algorithms. Enhancing images represent a fundamental task in
many image processing and vision applications. As a particularly
challenging case, restoring the weather degraded images requires
specic strategies and therefore, a variety of methods have been
emerged to solve this problem. Image enhancement methods tend
to increase the dynamic range and contrast of images degraded by
haze. This method does not need any scene depth information and
avoid complicated atmospheric scattering model.
In image processing area, the most commonly used nonmodel-based methods are histogram equalization and its variations
[1114]. Though histogram equalization method is simple and suitable for almost all types of images, but it fails to enhance the local
features. The local histogram equalization algorithm can process
the local lights feature very well, but large calculation and uncontrolled image quality are great disadvantages to it. So a substantial
amount of research focused on the improvement of the local histogram equalization algorithm. There are already many methods
have been proposed to fulll this requirement. Wadud et al. [15]
has proposed dynamic histogram equalization (DHE) for used in
the image contrast enhancement. Zhiyuan Xu et al. [16] present a
Contrast Limited Adaptive Histogram Equalization (CLAHE) based
method to remove fog from video sequences. The Major drawback
of this method is that, often, noisy pixels also get enhanced along
with enhancement of the foggy pixels. They also proposed a fog
degraded image contrast enhancement method based on Bilinear
Interpolation Dynamic Histogram Equalization (BIDHE). For color
images, histogram equalization can be applied to R, G, B color
channels separately. But this leads to undesirable change in hue.

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

Better results are obtained by rst converting the image to the

Hue, Saturation, Intensity color space and then applying histogram
equalization to the intensity component only [17]. However, even
this method does not fully maintain color delity.
Some other existing methods for enhancing the degraded
images, but these methods have some drawbacks. Standard ltering methods cannot restore images degraded by bad weather
conditions like fog, mist, haze, etc., hence contrast enhancement
methods are used. There are other, non-model-based methods
like unsharp masking [18], approaches based on the Retinex theory [1921] and wavelet-based methods [22,23]. Yu et al. [24]
has proposed a modied structure of the Retinex algorithm for
the image enhancement method. John and Wilscy [25] present
a novel method to enhance video sequences degraded by fog. In
this paper, background and foreground pixels are processed separately by a wavelet fusion method which reduces the computation
time. Generally, all non-model-based methods have a problem with
maintaining color delity. They also distort clear images, which is
one of the important limitation for fully automatic operation. There
are some shortages in traditional image enhancement methods of
fog image enhancement. For example, these methods cannot analyze the image comprehensively, it not only results in inevitably
noise amplication during its enhancement, but also introduces
new noises, even more may cause pseudo contour phenomenon.
But our method gives good visibility enhancement and also maintains the color delity.
3. Proposed method
To compensate for the limitations of these methods, a technique must be developed which creates a balance between high
levels of visual quality and low computational costs. In this paper,
a gamma correction using look-up-table methods is presented

1775

to accomplish this goal. As indicated in the description of the

histogram equalization, this is one of the most effective technique
for improving the contrast of the images. However, the traditional
histogram equalization method usually results in extreme contrast
enhancement, which causes an unnatural look and visual artifacts
present in the weather degraded images. Other techniques are
based on gamma adjustment method and thresholding techniques.
These techniques for enhancing the weather degraded images
improve the contrast of the images, but the processing time is
more. Our proposed method will improve the contrast of the
images as well as give less processing time in comparison to recent
reported algorithms [58]. Gamma correction techniques make up
a family of general histogram modications techniques obtained
simply by using a varying adaptive parameter . For various values
of  different levels of enhancements can be obtained. We present,
gamma correction using look-up-table method that improves the
contrast of the images and reduce the time complexity. Using a
look-up-table avoids needless repeated computations. The formula
to create a LUT for an 8-bit image is described as:
LUT = max intensity

 [0 : max intensity] 
max intensity

(1)

Fig. 2 shows the owchart of the proposed method for real time
deweathering of the images. In this approach the weather degraded
image is ltered using median lter. This lter is capable of reducing
noise and preserving the edges of the input image. The window size
of the median lter is used as 3 3 to have reduced processing time.
Edges are scaled by the constant gain factor of value 0.5.
Design procedure for the real time deweathering of the images
are carried out as below:
1. Take the weather degraded image.
2. Filtered weather images.

Fig. 2. Flowchart of the proposed algorithm.

1776

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

Fig. 3. (a), (b) and (c) Weather degraded images. (d), (e) and (f) Restored images by the proposed algorithm.

3. Edges extracted from the difference between the input image

and ltered image.
4. Constant gain factor multiply with the edges extracted from the
difference.
5. Perform image addition of the edge extracted and original image
to remove the artifacts present in the images.
6. Calculate the maximum pixel intensity present in the images.
7. The gamma correction is used for proper selection of gamma
value. Gamma correction allows users to better match the intensity.
8. Look-up-table method is used to get the nal enhanced image.
9. Output deweathering image.
The deweathering image recovered by the proposed approach
is shown in Fig. 3(d), (e) and (f). Our proposed method can enhance
a color image without generating halo artifacts or distorting the
color.
There are only few existing methods to enhance weather
degraded video sequences. However, these methods still consumes
considerable time, so that it is hard to satisfy real-time requirement
for some applications, especially for high video quality applications. This paper presents a fast approach to enhance the weather
degraded videos. Our method applies to a series of video frame
to enhance a video sequence. Fig. 4 shows the owchart of video
deweathering approach. At the beginning of the process, the rst
incoming frame is the processing frame of the video. Median lter
is used for ltering the videos. The size of the window is used as
3 3 which reduce the time complexity. Here simple frame subtraction has been used to extract the edges of the foreground part.
After that the edge extracted from the foreground part and previous frame is added to remove the artifacts present in the video. For
the weather degraded video used as input, most of the pixels are
densely distributed in the low-level region. So, gamma correction
method is used to enhance a video without generating artifacts or
distorting the color. At last, LUT method is used to get the nal
enhanced video. By applying our method, the processing rate is
greatly improved and color distortion is avoided.

Fig. 4. Video dehazing ow chart.

4. Comparison and experimental results

In order to demonstrate the effectiveness and robustness of our
proposed algorithm, we have experimented with our method as
well with best reported algorithms (Tan [5], Fattal [6], He et al.
[7], Tarel et al. [8] and Wang et al. [10]) on a large set of weather
degraded images and videos. Initially we have done the qualitative assessment extensively based on visual appearance for image
followed by video. Next we considered the important quantitative parameters in terms of blind contrast restoration assessment
[26] method (e, r), PSNR (peak signal to noise ratio) and processing
time for both image and also for video. Processing time is the time
taken by a processor to process an image or video using a particular algorithm. The time taken to process depends on the processor,
the simulator, image sizes and time complexity of the algorithm
itself. Processing time is an important performance metric while
considering a real time application. In the last part of quantitative

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

1777

Fig. 5. (a) Original color image, (b) Tan algorithm [5], (c) Fattal algorithm [6], (d) He et al. algorithm [7], (e) Tarel et al. [8], and (f) the proposed algorithm (from top to bottom
image name ny17, y01).

analysis, we have analyzed the effect of different size of images on

processing time and compared our approach with the best reported
fast algorithms i.e. Tarel et al. [8] and Wang et al. [10] method.
4.1. Qualitative comparison
In this section, we have compared with ve methods (Tan [5],
Fattal [6], He et al. [7], Tarel et al. [8] and Wang et al. [10]) in terms
of visual quality of the images and videos. Qualitative results of
selected deweathering algorithms are shown in Fig. 5. Fig. 5(a)
shows the original weather degraded color images followed by
enhanced images of the various dehazing algorithms. The dehazed
image obtained by Tan [5] is shown in Fig. 5(b). This algorithm is
applicable for both color and gray images. However, it does not
recover the original color of the scenes. Fig. 5(c) shows dehazed
image by Fattal [6] algorithm. Fattal [6] algorithm yields more natural results. However, this approach cannot well handle heavy haze
images. Again the Fattal [6] algorithm is based on the color information; hence, this method cannot be applied to the gray image. Fig. 5
(d) shows dehazed image by He et al. [7] algorithm. He et al. [7]
have introduced a novel dark channel prior method. This method

gives better results, but the processing time is high as transmission

map is rened by using soft matting. Fig. 5(e) shows dehazed image
by Tarel et al. [8] algorithm. Tarel et al. [8] proposed a fast visibility
restoration algorithm using minimum channel and median lter.
So, this algorithm is computationally less complicated and less
processing time as compared to other conventional method, but it
changes color tone and exhibits halo artifacts. The enhanced image
by our proposed method is given in Fig. 5(f). The visual analysis
of the enhanced images in Fig. 5(a)(f) indicates that the proposed
method gives better result in comparison to other proposed algorithm.
For comparing our dehazing approach for gray images, a gray
image is processed by other best algorithms and using our proposed
method. The performance of various approaches is shown in Fig. 6.
Fig. 6(a) shows original gray image. Fig. 6(b), (c) and (d) is dehazed
images by Tan [5], He et al. [7] and Tarel et al. [8], respectively.
These three algorithms are applicable for gray images. Fattal [6]
algorithm is not applied for grayscale images because this method is
basically a statistic interpretation that requires variance to estimate
the depth map. Fig. 6(e) is enhanced image by using our approach.
The visual analysis of the enhanced images in Fig. 6(a)(e) indicates

Fig. 6. (a) Original gray image, (b) Tan algorithm [5], (c) He et al. algorithm [7], (d) Tarel et al. [9], and (e) the proposed method.

1778

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

Fig. 7. (a) Original color image, (b) Tarel et al. [8], and (c) Wang et al. method [10], (d) the proposed method.

that the proposed method gives better result in comparison to other

proposed algorithm for gray image. So we can conclude that our
approach is suitable for both color and gray images.
Next, our approach is compared with Tarel et al. [8] and most
recent Wang et al. [10] method. Recently, Wang et al. [10] proposed a fast dehazing method which is based on a fusion and
image enhancement method. This method is slightly faster than
Tarel et al. [8] method. In Fig. 7(a) original images are shown.
The enhanced images by Tarel et al. [8] are shown in Fig. 7(b).
Fig. 7(c) shows dehazed image by Wang et al. [10] method. Fig. 7(d)
shows deweathering image by our proposed method. In Fig. 7, it is
observed that the yellow color sign board in road image is more
faithfully reconstructed using our method, similarly for cone and
stadium images.
After a thorough analysis of static images we worked on videos.
We have taken weather degraded videos, Riverside, and Road
View, sequences from Kims project page (http://mcl.korea.ac.kr/
projects/dehazing/) and applied our approach to it. To the best of
our knowledge, there is no algorithm specically given for foggy
and hazy videos. However, in prior work [8], it is mentioned that
the existing methods (Tan [5], Fattal [6], He et al. [7] and Tarel
et al. [8]) can be extended for video by making use of the temporal
correlation present among frames. We have extended the existing methods as well as our proposed approach for two standard
videos for fare comparison. First, we compare our approach with
Tan [5] method. Figs. 8(b) and 9(b) show the restored video frame
using Tan method. This method even though maximizing the contrast of the images but due to the overstretching of the contrast
it contains halo artifacts. We also compare our result with Fattal
[6] method. The results of the video frame using Fattal method are
shown in Figs. 8(c) and 9(c). This method gives more natural results,
but it cannot adequately remove haze and fog in some regions
which is visible in the restore frames. Next, in Figs. 8(d) and 9(d),
we compare the proposed method with He et al. [7] method. This

method depends only the dark pixel values, the information loss is
more if bright pixels are present in images. Tarel et al. [8] method
cannot completely eliminate the halos which can be seen from
Figs. 8(e) and 9(e). Figs. 8(f) and 9(f) show the restored deweathering, video frame using the proposed method. The proposed method
provides more reliable deweathering performance than existing
work. Our method diminishes halo artifacts and gives more natural results as compared with existing methods. The deweathering
results are appreciably improved by using the method which we
proposed. The processing speed of the video which can validate
that the proposed method is more suitable for real time processing
is discussed next.
4.2. Quantitative evaluation
For the quantitative analysis, blind contrast restoration assessment proposed by Hautieren et al. and PSNR is used. The blind
contrast restoration assessment method denes the ratio e of the
number of sets of new visible edges of the original image and the
restored image, as well as the average gradient ratio r. They are
expressed as:
e=

nr no
no

(2)

where nr and no are the numbers of visible edges in the original

image and the restored image.
r=

gr
go

(3)

gr and go are the average gradient of the original image and the
restored image. For haze removal, the greater e and r, the better
the dehazing effect. To determine the overall quality of the images,
PSNR indicator is used.
PSNR is a statistical measurement used for image or video quality assessment. PSNR is dened as the mean square error (MSE) of

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

1779

Fig. 8. Video deweathering on the Riverside sequence. (a) Three frames in a foggy video, (b) The result using Tan method, (c) The result using Fattal method, (d) The result
using He et al. method, (e) The result using Tarel et al. method and (f) The result using the proposed method. From top to bottom, the frame numbers are 10, 12, and 14,
respectively.

two M N images x and x , where one of the image is considered a

noisy counterpart of the other.
The PSNR is calculated as follows.
(2s 1)
PSNR = 10 log
MSE

where MSE is given by:

(4)

2
1 
xij xij ,
MN
M1 N1

MSE =

0 i M 1,

0j N1

i=0 j=0

(5)

Fig. 9. Video deweathering on the Road View sequence. (a) Three frames in a foggy video, (b) The result using Tan method, (c) The result using Fattal method, (d) The
result using He et al. method, (e) The result using Tarel et al. method and (f) The result using the proposed method. From top to bottom, the frame numbers are 7, 9, and 11,
respectively.

1780

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

Table 1
Quantitative and processing time comparison.
Image

Method

PSNR

Time/second

ny17

Tan
Fattal
He et al.
Tarel et al.
Proposed method

0.0356
0.0340
0.0155
0.0139
0.1056

2.0234
1.4094
1.5448
1.5311
1.1879

59.0426
69.2153
66.0207
59.1575
56.0229

19.79
2.92
28.79
2.80
0.88

y01

Tan
Fattal
He et al.
Tarel et al.
Proposed method

0.0190
0.0171
0.0141
0.0098
0.0264

2.2723
1.1682
1.3740
1.8463
1.1749

56.7941
66.5738
58.9532
63.9328
59.2802

19.79
2.92
28.79
2.80
0.73

Road image

Tan
He et al.
Tarel et al.
Proposed method

0.838
0.7060
0.5875
0.7604

1.8976
1.7915
2.7153
1.917

62.0133
58.4418
66.2975
59.0297

9.76
20.16
2.10
0.45

where xij is the original image, xij is the dehazed image, M and N
represent the image size, (2s 1) corresponds to the theoretical
maximum intensity value of the image and s is the depth of the
image (s = 8). Tables 1 and 2 show the comparison results of the
existing methods and proposed method in terms of e, r, PSNR value
and processing time. In Table 1 three standard images (i.e. ny17, y01
and Road image) and in Table 2 two standard videos (Riverside,
and Road View,) are considered. Here, our method is compared
with Tan [5], Fattal [6], He et al. [7] and Tarel et al. [8]. Generally,
a larger PSNR values implies a better visual quality of the images.
In our case, comparable PSNR results along with e and r show the
enhanced contrast deweathering image without losing any visual
appearance. The resultants are shown in Figs. 5(f), 6(e), 8(f) and 9(f),
respectively, where it is evident that the proposed approach is capable of enhancing the quality of the images. Here, Figs. 8(f) and 9(f)
show the video deweathering frames using our approach. Furthermore, our algorithm achieves higher PSNR values for these videos.
If the excessive contrast is present in the restored video frame by
Tan [5], the color of the fog free is often oversaturated and looks
unnatural, e and r value will be very high and artifacts are present
in the recovered images and video frames.
Next, based on processing time, we can analyze how much our
algorithm is faster than existing works. Analyzing the results of
Table 1, Tan [5] and He et al. [7] method need more than 10 and
20 second per image. Fattal [6] method needs more than 2 second
for processing the images. Tarel et al. [8] method takes 2 second

per image. It is clear that the computational complexity of the proposed method is very less along with enhance the contrast of the
restored image. Computational complexity plays an important role
for real time processing and this complexity varies with different
image sizes. On the basis of these results, we can conclude that the
proposed algorithm is faster than existing methods.
For obtaining average processing time, the same two standard
videos of 30 frames per second are taken. The existing methods as
well as our proposed approach are applied to both the videos. The
processing time obtained for both the videos of one second duration consisting of 30 frames are listed in Table 2. The result shows,
Tan [5] algorithm takes almost 1415 second for processing one
second duration video. So, this method is not suitable for real time
processing. Fattal [6] method also takes 810 second for processing
the average of 30 frames. He et al. [7] approach is computationally
expensive. Tarel et al. [8] method is computationally less complicated and a good candidate for the video application in future. But
in comparison with this method, our approach is taking almost
half the processing time. This results approves that the signicant
improvement in speed of the proposed algorithm. Thus, the proposed method is more suitable for real time implementation of the
fog and haze removal of restored videos without losing any visual
quality.
Finally, we have compared e, r, PSNR value and processing
time with the best reported dehazing algorithms such as Tarel
et al. [8] and Wang et al. [10] method by simulating all of them

Table 2
Quantitative estimation of video deweathering.
Weather degraded input
sequence

Tan

Fattal

PSNR

Time (Sec.Per
30 frames)

He et al.

PSNR

Time (Sec.Per
30 frames)

PSNR

Time (Sec.Per
30 frames)

River side

Frame no. 10
Frame no. 12
Frame no. 14

0.1051
0.1378
0.1652

2.1055
2.1052
2.1059

59.3692
59.3716
59.3627

14.4536

1.0954
0.0995
0.1302

1.2103
1.9499
1.9433

55.5941
55.5579
55.5454

8.18442

0.1308
1.5689
0.1239

1.4110
0.6947
1.4930

58.2038
55.9097
58.7705

740.662

Road view

Frame no. 7
Frame no. 9
Frame no. 11

1.3857
1.2680
1

1.4352
1.4364
1.4641

56.6065
56.6223
56.6105

15.9270

0.1096
0.0850
0.1616

1.2681
1.2895
1.2875

56.2881
57.8902
58.0679

10.6871

0.9967
0.9967
0.2195

1.0211
0.8449
0.9759

57.1285
58.2099
57.6657

852.713

Weather degraded input

sequence

Tarel et al.

Proposed method

PSNR

Time (Sec.Per
30 frames)

River side

Frame no. 10
Frame no. 12
Frame no. 14

0.0318
0.0587
0.0638

1.8352
1.8391
1.8464

61.5819
61.5809
61.5586

3.4220

Road view

Frame no. 7
Frame no. 9
Frame no. 11

0.8140
0.8268
0.6128

1.3918
1.4023
1.4075

56.1858
56.1966
56.2080

4.879

0.0857
0.0756
0.0700

1.4414
1.4114
1.3231

64.2456
64.3277
64.2390

1.775

0.3555
0.2876
0.0152

0.9679
0.9599
0.9606

60.2669
60.2392
60.2593

2.270

PSNR

Time (Sec.Per
30 frames)

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

1781

Table 3
Quantitative and processing time comparison (in sec) with Tarel [8] and Wang [10] for different image sizes.
Image sizes

Tarel

Wang

465 384 (s)

0.1526
600 400 (s) 0.0352
1000 327 (s) 0.0756

Our

PSNR

Time/
second

PSNR

Time/
second

PSNR

Time/
second

1.6309
1.7964
1.6490

58.4344
60.4633
59.4904

1.57
2.08
2.86

0.0387
0.0313
0.2854

1.9853
2.1797
1.704

58.0774
59.5027
58.6550

1.46
1.87
2.49

0.0752
0.0196
0.0244

1.3257
1.2436
1.3835

58.7848
57.6408
59.7815

0.659
0.997
1.159

on the same hardware platform. Our method gets the comparable

results as compared with [8,10] in terms of visual quality. The
result obtained by Tarel et al. [8] and Wang et al. [10] methods
are compared and analyzed with the help of the comparison chart.
For fare comparison, three images of size 465 384, 600 400 and
1000 327 are taken as input original image. The processing time
for different image sizes using Tarel et al. [8] and Wang et al. [10]
method is obtained as 1.57 s, 2.08 s, 2.86 s and 1.46 s, 1.87 s, 2.49 s,
respectively. The processing time for our method for same three
gures are obtained as 0.659 s, 0.997 s and 1.159 s. So as compared
with Tarel et al. [8] and Wang et al. [10], our method takes less
processing time.

% improvement
w.r.t. Tarel

% improvement
w.r.t. Wang

41.9
48.0
41

45.1
53.3
46.5

image method to be applied in real-time processing applications.

In this paper, we apply a median lter and gamma correction
using look-up-table method to enhance the weather degraded
images. Our approach is giving a better result with speed improvement up to 48 % and 53 % as compared with the references [8]
and [10]. Experimental results demonstrated that the proposed
algorithm has a signicant effect on deweathering, as well as
achieving real-time video processing. The proposed method with
less processing time, may be useful for many systems ranging from
surveillance, intelligent vehicles, to remote sensing, etc. Future
research work will be focused on hardware implementation of this
method.
Tarel
Wang
Our

Comparison chart of Image Sizes v/sProcessing time.

3
2.5

1.5
1

time

Processing

0.5
0

465 x 384

600 x 400
Image Sizes

For a quantitative comparison, the speed improvement over

Tarel and Wang are obtained and summarized in the last two
columns of Table 3. It is found that the proposed approach is faster
by up to 48% and 53% in comparison to Tarel and Wang methods respectively. The speed improvement of our method is better
because of the use of only one median lter and look-up-table
method which take less computation time as compared to Tarel
et al. [8] and Wang et al. [10] method. Tarel et al. [8] need to apply
a median ltering on the median. This means that their method
must be carried out by two median lters. So the processing time
of Tarel et al. [8] is more than our computational time. The computational complexity of Wang et al. [10] method is the minimum
lter approach.
The processing time is less than Tarel et al. [8] method and it is
almost close to it. Therefore, our method is better than Tarel et al.
[8] and Wang et al. [10].
5. Conclusion
In this paper, we propose a new method for enhancing the
image and video by using look-up-table method. The main contribution of our method is less processing time as compared with
other existing methods without sacricing the enhanced quality.
This less processing time allows the enhanced weather degraded

1000 x 327

References
[1] Narasimhan S, Nayar S. Vision and the atmosphere. Int J Comput Vis
2002;48(3):23354.
[2] Narasimhan S, Nayar S. Contrast restoration of weather degraded images. IEEE
Trans Pattern Anal Mach 2003;25(6):71324.
[3] Shwartz S, Namer E, Schechner Y. Blind haze separation. In: Proc. IEEE CVPR.
2006. p. 198491.
[4] Schechner Y, Narasimhan S, Nayar S. Instant dehazing of images using polarization. In: Proc. IEEE CVPR. 2001. p. 32532.
[5] Tan R. Visibility in bad weather from a single image. In: Proc. IEEE CVPR. 2008.
p. 18.
[6] Fattal R. Single image dehazing. ACM Trans Graph 2008;27(3):19.
[7] He K, Sun J, Tang X. Single image haze removal using dark channel prior. IEEE
Trans Pattern Anal Mach 2011;33(12):195663.
[8] Tarel J, Hautire N. Fast visibility restoration from a single color or gray level
image. In: Proc. IEEE ICCV. 2009. p. 22018.
[9] He K, Sun J, Tang X. Guided image ltering. In: IEEE Trans. Pattern Anal. Mach.
Intell. 2013. p. 1397409.
[10] Wang Z, Feng Y. Fast Single Haze Image Enhancement. Computers and Electrical
Engineering; 2013.
[11] Pizer SM, Philip Amburn E, Austin JD, Cromartie R, Geselowitz A, Greer T, et al.
Adaptive histogram equalization and its variations. Comput Vis Graph Image
Process 1987;39:35568.
[12] Zuiderveld K. Contrast limited adaptive histogram equalization. In: Heckbert
P, editor. Graphics Gems IV. New York: Academic; 1994. p. 47485 [Chapter
VIII.5].

1782

A. Kumari, S.K. Sahoo / Int. J. Electron. Commun. (AE) 69 (2015) 17731782

[13] Stark JA. Adaptive image contrast enhancement using generalizations of histogram equalization. IEEE Trans Image Process 2000 May;9(5):88996.
[14] Rodriguez JJ, Yang CC. High-resolution histogram modication of color images.
Graph Models Image Process 1995 Sep;57(5):43240.
[15] A-Al-Wadud M, Hasanul Kabir Md, Ali Akber Dewan M, Chae O. A dynamic
histogram equalization for image contrast enhancement. IEEE Trans Consum
Electron 2007;53:593600.
[16] Xu Z, Liu X, Chen X. Fog removal from video sequences using contrast limited
adaptive histogram equalization. In: CISE, International Conference on, IEEE.
2009. p. 14.
[17] Rodriguez JJ, Yang CC. High-resolution histogram modication of color images.
Graph Models Image Process 1995 Sep;57(5):43240.
[18] Polesel A, Ramponi G, Mathews VJ. Image enhancement via adaptive unsharp
masking. IEEE Trans Image Process 2000;9(March (3)):50510.
[19] Jobson DJ, Rahman Z, Woodell GA. Properties and performance of a center/surround retinex. IEEE Trans Image Process 1997;6(March (3)):45162.

[20] Jobson DJ, Rahman Z, Woodell GA. A multi-scale retinex for bridging the gap
between color images and the human observation of scenes. IEEE Trans Image
Process 1997;6(July (7)):96576.
[21] Rahman Z, Jobson DJ, Woodell GA. Retinex processing for automatic image
enhancement. J Electron Imag 2004;13(January (1)):10010.
[22] Scheunders P. A multivalued image wavelet representation based on multiscale
fundamental forms. IEEE Trans Image Process 2002;10(May (5)):56875.
[23] Grewe L, Brooks RR. Atmospheric attenuation reduction through multi-sensor
fusion. Proc SPIE 1998;3376:1029.
[24] Yu B-N, Kim BS, Lee KH. Visibility enhancement based real-time retinex for
diverse environments. SITIS 2012.
[25] John Jisha, Wilscy M. Enhancement of weather degraded video sequences using
wavelet fusion. In: CIS, IEEE Conf. 2008. p. 16.
[26] Hautiere N, Tarel J-P, Aubert D, Dumont E. Blind contrast enhancement
assessment by gradient ratioing at visible edges. J Image Anal Stereol
2008;27(2):8795.