A Review On Video and Image Defogging Algorithms Image Restoration Enhancement

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

International Journal of Engineering and Technical Research (IJETR)

ISSN: 2321-0869 (O) 2454-4698 (P), Volume-5, Issue-1, May 2016

A Review on Video and Image Defogging Algorithms


Image Restoration Enhancement
Gurjeevan Kaur, Jashanpreet Kaur

5].Because the existing defogging or dehazing algorithms


Abstract In this paper, we propose a simple but effective have no clear boundaries, in this we use image defogging to
image prior - dark channel prior to remove haze from a single refer to algorithms that have the ability to remove fog or haze
input image. The dark channel prior is a kind of statistics of the from the image. Many improved defogging algorithms based
haze-free outdoor images. It is based on a key observation - most
on the physical model were proposed for outdoor scenes.
local patches in haze-free outdoor images contain some pixels
which have very low intensities in at least one color channel.
Some video and image defogging algorithms were also
Using this prior with the haze imaging model, we can directly proposed for real-world traffic surveillance scenes [6-9]. In
estimate the thickness of the haze and recover a high quality order to improve the visibility of the unmanned surface
haze-free image. Another downside is that the captured hazy vehicle (USV) visual system, Ma et al. presented an improved
road image contains localized lightweight sources or color-shift image defogging algorithm based on the dark channel prior to
issues owing to duster conditions. the target of this work is to foggy sea image restoration, and the proposed single image
implement the Road Scenes Captured by Intelligent defogging algorithm was also extended to fast video
Transportation Systems mistreatment Hybrid technique. to defogging [10]. Under poor visibility conditions such as
reinforce the pictures mistreatment completely different filters,
foggy weather, it is difficult to find runways and hazards from
restoration and improvement techniques.
the visual system of a flight. In order to solve this problem, a
Index Terms Scene restoration , Depth estimation, Hazy contrast enhancement algorithm based on the Retinex theory
Images , Scene radiance, Transmission etc. which can effectively improve the visibility of the foggy
image acquired by the flight visual system [11]. Most existing
I. INTRODUCTION defogging algorithms also aim at removing fog from land
images. However, there are few studies of the sea and air. The
Haze and fog are a typical marvel ashore and sea. In foggy and
above applications all demonstrate that the video and image
dim climate, there are numerous environmental particles of
defogging algorithms are significant and well worth studying.
noteworthy size. They not just assimilate and dissipate the
reflected light of the scene, additionally disseminate some
environmental light to the camera. In this manner, the picture
II. DETECTION AND CLASSIFICATION OF FOGGY
procured by the camera is debased and for the most part has
IMAGES
low complexity and poor perceivability [1]. This will truly
impact the visual framework particularly the unmistakable Existing video and picture defogging calculations are
light visual framework. Because of the debasement of the dependably specifically connected to the picture paying little
picture, the objectives and hindrances of the picture are hard heed to the vicinity or nonappearance of mist. Be that as it
to distinguish. This is terrible for robotized video preparing, may, for true applications, it is important to know whether the
for example, highlight extraction, target following, and picture gained in the present environment should be prepared
acknowledgment of articles. This is additionally one of the by a defogging calculation. The fundamental reason is as per
principle purposes behind mischance noticeable all around, the following: perceivability of the restored picture acquired
on the ocean, and out and about. So it is vital to plan a picture by the defogging calculation might be more regrettable than
defogging calculation to enhance the natural versatility of the the first picture if no judgment is made. Additionally, the
visual framework. With the advancement of PC innovation, utilization of the defogging calculation is tedious, which is not
the video and picture defogging calculations have gotten gainful to understand the constant target identification,
much consideration and are generally connected in common following, and acknowledgment. There are two strategies
and military fields, for example, remote detecting, target which can judge whether the present scene has mist or not.
recognition, and activity reconnaissance. [2] utilized the The principal system is the mist recognition strategy which
picture defogging calculation to improve the perceivability of respects the undetectable zone of the picture as the foggy
the vehicle visual framework, which can viably counteract region. The second strategy is the foggy picture grouping
auto collisions. For pictures of outside scenes, examined the technique. A. Discovery of foggy zones in pictures Two
visual indications of various climate conditions, for example, techniques can recognize the foggy zones of the picture. The
dimness, mist, cloud, and rain [3], and after that built up a primary strategy depends on the semi-reverse picture, and the
physical imaging model taking into account the barometrical second technique depends on the meteorological
disseminating wonder for picture defogging [4, perceivability separation. 1) Foggy territory location in light
of the semi-converse picture initially proposed a foggy region
discovery calculation in view of the 'semi-backwards' picture
5 [3]. The semi-reverse picture S is acquired by selecting the
Jashanpreet Kaur, Assistant Professor, Guru Kashi University, Talwandi most extreme of the first picture pixel and its converse picture
Sabo pixel which is defined as
Gurjeevan Kaur, M.Tech, Research Scholar, Guru Kashi University,
Talwandi Sabo

80 www.erpublication.org
A Review on Video and Image Defogging Algorithms Image Restoration Enhancement

line to particular the foggy range and clear territory. In any


case, the meteorological perceivability separation can be
where c denotes one of the RGB channels, I is the original
utilized to judge which zone has dainty mist or thick mist.
image, and 1 -Ic denotes the inverse image of the original
This might be useful to appraise the parameters of defogging
image. After renormalizing the inverse image, detected the
calculations.
foggy areas in the channel of the Lch color space, and regarded
the pixels which have a large difference between the
III. IMAGE DEFOGGING BASED ON IMAGE
semi-inverse image and original image as clear pixels, and
RESTORATION
regarded the remaining pixels as foggy pixels. This foggy area
detection method is based on the fact that the intensity values In 1976, McCartney first proposed the atmospheric scattering
of pixels in the foggy area of the image are usually much physical model based on the Mie scattering theory [5]. Fig.1
bigger than those of pixels in the clear area. In the sky or shows the physical atmospheric scattering model under foggy
foggy areas of an image, pixels usually have a high intensity in weather. The physical model is composed of the airlight
all color channels. Thus, the semi-inverse image will have the model and direct transmission model. Fig.4 also shows the
same value as the original image in these areas. However, in degradation reason of images under foggy conditions. In the
clear areas, there is at least one channel of the semi-inverse direct transmission model, the light for imaging will be
image where pixel values will be replaced by the inverse attenuated by atmospheric scattering, which leads to the
image. In other words, the output of Eq.(1) is respectively the degradation of edge details and object textures of the image.
original image in foggy areas and the inverse image in clear In the airlight model, some sunlight will also be scattered by
areas. Then the foggy area can be easily detected by the the atmosphere and transmitted to the camera, and these lights
difference between the original image and its semi-inverse are not the scene lights and can be considered as the fog
image. This algorithm is simple and effective for detection of component of the image whose influence is similar to that of a
foggy areas in foggy images, but it is not suitable for the veil to hide the objects in the image. For a clear image, the
judgment of whether the current scene has fog or not. This is direct transmission model makes up a large proportion in the
because the sky area or white area of a clear image will be imaging model. With an increase of the concentration of the
mistaken for a foggy area via this algorithm. fog, the proportion of the direct transmission model will
decrease while the proportion of the airlight model will
2) Foggy area detection based on the meteorological increase and visibility of the image will decrease. In other
visibility distance words, the airlight model is the main reason that leads to an
The International Commission on Illumination (CIE) image acquired under foggy conditions being a fuzzy image
characterized the meteorological perceivability separation of with low contrast and visibility. Regarded that the scattering
a picture and its estimation strategy [2]. The meteorological coefficient is not relevant to the wavelength of visible light in
perceivability separation of a picture is generally connected in the homogenous atmosphere and presented a simplified
the field of foggy picture recognition of a vehicle visual physical model for image restoration:
framework [3].
For a foggy picture, proposed a daytime foggy range
recognition calculation by means of figuring the
meteorological perceivability separation [4]. They initially
utilized the Canny-Deriche channel to separate the picture
forms in order to highlight the edges of roadways. At that
point the locale developing calculation was performed to
discover the street surface layer. Third, they set up four
conditions to acquire the objective locale. At long last, the
perceivability separation of the picture was gotten by
ascertaining the estimation data transfer capacity. An even
line to signify the perceivability separation. For the vehicle
camera framework, the locale over the even line more often
than not has low difference and can be viewed as the foggy
zone or undetectable range. Bronte et al. likewise
distinguished the foggy region of a picture by means of Fig.1 Physical atmospheric scattering under foggy conditions
evaluating the perceivability separation [5].
The mist discovery technique in view of the meteorological I. LITERATURE SURVEY
perceivability separation partitions the foggy picture into two
locales: unmistakable range and undetectable territory. In I studied so many papers and some of them are given below:
spite of the fact that the haze recognition calculation can
distinguish the foggy region of pictures, it likewise has a few Yong Xu et.al. [2014] have presented Video and images
inadequacies. The undetectable region over the level line of acquired by a visual system are seriously degraded under hazy
the picture does not implies that it ought to be completely and foggy weather, which will affect the detection, tracking,
doled out to the foggy territory. Some removed scenes of and recognition of targets. Thus, restoring the true scene from
common clear pictures additionally look hazy and might be such a foggy video or image is of significance. The main goal
mixed up for undetectable regions or foggy regions by the of this paper was to summarize current video and image
mist location calculation. Besides, for some foggy pictures defogging algorithms. We first presented a review of the
with inhomogeneous haze circulation, it is elusive an even detection and classification method of a foggy image. Then,

81 www.erpublication.org
International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869 (O) 2454-4698 (P), Volume-5, Issue-1, May 2016
we summarized existing image defogging algorithms and natural elements causing obstruction, such as foliage
including image restoration algorithms, image contrast growing into the scene and ice forming on the porthole. In
enhancement algorithms, and fusion-based defogging addition, other persistent view changes resulting from new
algorithms. We also presented current video defogging scene elements of interest being captured, such as stalled cars,
algorithms. We summarized objective image quality suspicious packages, etc. might warrant alarms. Since these
assessment methods that have been widely used for the systems are often unattended, it is often important to
comparison of different defogging algorithms, followed by an automatically detect such incidents early.[3].
experimental comparison of various classical image
defogging algorithms. Finally, we presented problems of Fan-Chieh Cheng et.al [2011] have proposed a novel
video and image defogging which need to be further background subtraction approach in order to accurately detect
studied.[5] moving objects. The method involves three important
proposed modules: a block alarm module, a background
Shih-Chia Huang et.al [ 2014] have presents The visibility modeling module, and an object extraction module. The block
of images of outdoor road scenes will generally become alarm module efficiently checks each block for the presence
degraded when captured during inclement weather of either a moving object or background information. This is
conditions. Drivers often turn on the headlights of their accomplished by using temporal differencing pixels of the
vehicles and streetlights are often activated, resulting in Laplacian distribution model and allows the subsequent
localized light sources in images capturing road scenes in background modeling module to process only those blocks t
these conditions. Additionally, sandstorms are also weather at were found to contain background pixels. Next, the
events that are commonly encountered when driving in some background modeling module
regions. In sandstorms, atmospheric sand has a propensity to is employed in order to generate a high-quality adaptive
irregularly absorb specific portions of a spectrum, thereby background model using a unique two-stage training
causing color-shift problems in the captured image. procedure and a novel mechanism for recognizing changes in
Traditional state-of-the-art restoration techniques are unable illumination. The overall results show that our proposed
to effectively cope with these hazy road images that feature method attains a substantially higher degree of efficacy,
localized light sources or color-shift problems. In response, outperforming other state-of-the-art methods by Similarity
we present a novel and effective haze removal approach to and F1 accuracy rates of up to 35.50% and 26.09%,
remedy problems caused by localized light sources and color respectively. [4].
shifts, which thereby achieves superior restoration results for
single hazy images. The performance of the proposed method IV. PROBLEM FORMULATION
has been proven through quantitative and qualitative The problem undertaken for the dissertation is a novel and
evaluations. [1]. effective haze removal approach to remedy problems caused
by localized light sources and color shifts, which thereby
Ivan et.al [2014] have studied work deals with multi-label achieves superior restoration results for single hazy images.
classification of traffic scene images. We introduce a novel The Road image degradation can cause problems for
labeling scheme for the traffic scene dataset FM2. Each image intelligent transportation systems such as traveling vehicle
in the dataset is assigned up to five labels: settlement, road, data recorders and traffic surveillance systems, which must
tunnel, traffic and overpass. They propose representing the operate under a wide range of weather conditions. Another
images with (i) bag-of-words and (ii) GIST descriptors. The problem is that the captured hazy road image contains
bag-of-words model detects SIFT features in training images, localized light sources or color-shift problems due to
clusters them to form visual words, and then represents each sandstorm conditions. Motion detection is known to be one of
image as a histogram of visual words. On the other hand, the the greatest problem areas. There is darkness problem in the
GIST descriptor represents an image by capturing perceptual road scene images due to sandstorms.
features meaningful to a human observer, such as naturalness,
openness, roughness, etc. We compare the two V. METHODOLOGY
representations by measuring classification performance of
Support Vector Machine and Random Forest classifiers. The dissertation is to improvement in Road Scenes Captured
Labels are assigned by applying binary one-vs-all classifiers by Intelligent Transportation Systems. It is based upon GUI
trained separately for each class. Categorization success is (graphical user interface) in MATLAB. It is an effort to
evaluated over multiple labels using a variety of parameters. further grasp the fundamentals of MATLAB and validate
They report good classification results for easier class labels it as a powerful application tool. There are basically
(road, F1 = 98% and tunnel, F1 = 94%), and discuss weaker different files. Each of them consists of m-file and figure file.
results (overpass, F1 < 50%) that call for use of more An effective approach for the haze removal of single images
advanced methods. [2]. captured during different environmental conditions that not
only avoids the generation of artifact effects but also recovers
Ajay Raghavan et.al [2012] have presented Unattended true color. Our approach involves three proposed modules,
camera devices are increasingly being used in various i.e., an HDCP module, a RGB module, and a VR module and
intelligent transportation systems (ITS) for applications such enhancement techniques and filters. The proposed HDCP
as surveillance, toll collection, and photo enforcement. In module designs an effective transmission map to circumvent
these fielded systems, a variety of factors can cause camera halo effects in the recovered image and estimates the location
obstructions and persistent view changes that may adversely of the atmospheric light to avoid underexposure. In order to
affect their performance. Examples include camera recover the true color of scenes featuring a wide range of
misalignment, intentional blockage resulting from vandalism, weather conditions, we propose the RGB module. This RGB

82 www.erpublication.org
A Review on Video and Image Defogging Algorithms Image Restoration Enhancement

module determines the intensity statistics for the RGB color [6] J.-E. Ha et.al , Foreground objects detection using multiple difference
images, Opt. Eng., vol. 49, no. 4, p. 047 201,Apr. 2010.
space of a captured image in order to acquire the color
[7] [F. Barranco et.al Visual system based onartificial retina for motion
information. detection, IEEE Trans. Syst., Man, Cybern.B, Cybern., vol. 39, no. 3,
pp. 752762, Jun. 2009.
[8] A. Doshi et.al Smoothing of optical flow using robustifieddiffusion
kernels, Image Vis. Comput., vol. 28, no. 12, pp. 15751589,Dec.
2010.
[9] C. Stauffer et.al Adaptive background mixture modelsfor real-time
tracking, in Proc. IEEE Comput. Vis. Pattern Recog., 1999,pp.
246252.

Fig.4.1 The categories of video algorithms

The following steps are proposed for this work:


Step 1: Read the image and video that includes the road
scenes.
Step 2: Apply the preprocessing technique to process the
image and video.
Step 3: Apply the hybrid technique to enhance the road
scenes.
Step 4: Apply the enhancement technique to enhance the
image and video and road scenes.
Step 5: Remove the darkness of the images and video.
Step 6: repeat the step for multiple road scenes.
Step7: calculate the parameters.
Step 8: Stop.

VI. CONCLUSION
In this paper, we have proposed a very simple but powerful
prior, called dark channel prior, for single image haze
removal. The dark channel prior is based on the statistics of
the outdoor images. Applying the prior into the haze imaging
model, single image haze removal becomes simpler and more
effective. In this paper totally different papers are studied by
the various researchers to boost the videos and therefore the
pictures of the road scenes. We tend to summarized existing
image defogging algorithms as well as image restoration
algorithms, image distinction sweetening algorithms, and
fusion-based defogging algorithms. We tend to additionally
given current video defogging algorithms. Our work also
shares the common limitation of most haze removal methods -
the haze imaging model may be invalid.

REFERENCES

[1] Shih-Chia Huang et.al An Efficient Visibility Enhancement Algorithm


for Road Scenes Captured by Intelligent Transportation Systems IEEE
Transactions On Intelligent Transportation Systems, Vol. 15, No. 5,
OCTOBER 2014.
[2] Ivan et.al Multi-Label Classification of Traffic Scenes Proceedings of
the Croatian Computer Vision Workshop, Year 2 September 16, 2014,.
[3] Ajay Raghavan et.al Detection of Scene Obstructions and Persistent
View Changes in Transportation Camera Systems 2012 15th
International IEEE Conference on Intelligent Transportation
SystemsAnchorage, Alaska, USA, September 16-19, 2012.
[4] Fan-Chieh Cheng et.al Accurate Motion Detection Using a
Self-Adaptive Background Matching Framework IEEE Transactions
On Intelligent Transportation Systems, Vol. 13, No. 2, JUNE 2012
[5] Yong Xu et.al. Review of video and image defogging algorithms and
related studies on image restoration and enhancement Translations and
content mining are permitted for academic research only. Personal use is
also permitted, but republication/redistribution requires IEEE
permission IEEE 2015.

83 www.erpublication.org

You might also like