Seminar Report(Chiranjivi)

7/23/2019 Seminar Report(Chiranjivi)

1/22

COCHIN UNIVERSITY OF SCIENCE &TECHNOLOGY

KOCHI - 682022

SEMINAR REPORT

on

UNDERWATER IMAGE ENHANCEMENT

BY DARK PRIOR CHANNEL

In partial fulfillment for the award of the degree

Of

Bachelor Of Technology

In

ELECTRONICS &COMMUNICATION

ENGINEERING

at

COCHIN UNIVERSITY COLLEGE OF ENGINEERINGKUTTANAD

Submitted by

CHIRANJIVI KUMAR(Reg.no. 13130216)


2/22

COCHIN UNIVERSITY OF SCIENCE &TECHNOLOGY

COCHIN UNIVERSITY COLLEGE OF ENGINEERING

KUTTANAD- 688504

Department of

ELECTRONICS AND COMMUNIACTION ENGINEERING

CERTIFICATE

Certified that this is a bonafied record of the seminar entitled

UNDERWATER IMAGE ENHANCEMENTBY DARK PRIOR CHANNEL

Submitted by

CHIRANJIVI KUMAR(reg. no.13130216)

During the year 2015 in partial fulfillment of the requirements for the award of

Degree ofBachelor of Technology inElectronics & communication Engineering

of Cochin University of Science & Technology.

Seminar co-ordinator Head of the Department

Mr. SAM VARGHESE Mr. ANIL KUMAR K K


3/22

ACKNOWLEDGEMENT

Many people have contributed to the success of this paper. Although a single sentence hardly

suffices, I would like to thank the Almighty God for blessing me with his grace and taking my

endeavor to a successful culmination. I am extremely grateful to my Principal

Dr. P.S Srijith for his wholehearted cooperation for the successful completion of paper. I

extend my sincere and heart-full thanks to Mr. Anil Kumar K.K. Head of Department

(Electronics And Communication Engineering) for providing me the right ambience forcarrying out the work on this paper and for the facilities provided to me. I am profoundly

indebted to my seminar- in charge, Mrs. Akhila L,lecturer in Department of Electronics And

Communication Engineering for her innumerable acts of timely advice, encouragement and I

sincerely express my gratitude to her. I would like to extend my gratitude to all the staff of the

Department of Electronics And Communication Engineering for the help and support rendered

to me. I have been benefitted a lot from the feedback and suggestions given to me by them.

CHIRANJIVI KUMAR

(Reg.no. 13130216)


4/22

ABSTRACT

Light scattering and color change are two main problems in underwater images. Due to light

scattering, incident light gets reflected and deflected multiple times by particles present in the

water. This degrades the visibility and contrast of the underwater image.

Dark channel prior is method used for removing the haze present

in the underwater image. It is based on a key observation - most local patches in haze-free

underwater images contain some pixels which have very low intensities in at least one color

channel. Using this prior with the haze imaging color model estimates the thickness of the

haze and recover a high quality haze free image.

This method does not require images with different exposure

values, and is entirely based on the attenuation experienced by point across multiple frames.

In this paper underwater image enhancement using Dark channel prior is attempted. Results

shows that the performance of this method is better compared to image enhancement using

histogram equalization.

The existing research shows that underwater images give new

challenges and significant problems because of light scattering effect. From the last few

years, a growing interest in marine research has encouraged researchers from different

disciplines to explorer the mysterious underwater world. A significant amount of literature

is available on underwater image processing.


5/22

Underwater Image Enhancement

DEPT.OF ECE 1 CUCEK

SEPTEMBER 2015

TABLE OF CONTENTS

1. Introduction

What Is Image Enhancement 1

2. Types of image Enhancement 2

2.1Histogram Equalization

2.2

Image Dehazing

2.3WCID Process

3. Dark Prior Channel 11

3.1Color feature Extraction

3.2

Estimating the light source of illumination

3.3

Dark Channel Estimation

3.4Haze Removal

4. Experimental Results 15

5. Conclusion 17

6. References 18


6/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

2

LIST OF FIGURES

FIGURE NO. TITLE PAGE NO.

1.

Natural light enters in the water from air 3

2. Histogram equalization applied to low 5

Contrast image

3. The original image and dehazed image in Fog 7

4. Image formation model 8

5.

WCID flow diagram 9

6.

Underwater enhanced Image 15


7/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

3

1. INTRODUCTION

Image enhancement improves the visibility of one aspect or component of an image. It

refers to sharpening of image features such as boundaries, or contrast to make a graphic

display. This is mainly useful for display & analysis. This process will not increase the

intrinsic information of an image. This includes gray level & contrast manipulation, noise

reduction, sharpening, filtering, interpolation and magnification, pseudo coloring, and so on.

Image enhancement uses qualitative subjective approach to produce a more visually

pleasing image. Acquiring clear images in underwater environments is an important issue in

ocean engineering.

The quality of underwater images plays a pivotal role in scientific

missions such as monitoring sea life, taking census of populations, and assessing geological

or biological environments. Capturing images underwater is challenging, mostly due to haze

caused by light that is reflected from a surface and is deflected and scattered by water

particles. Due to varying degrees of attenuation encountered by different wavelengths of

light, underwater images always dominated by a bluish tone. Light scattering and color

change result in contrast loss and color deviation in images acquired underwater. Haze is

caused by suspended particles such as sand, minerals, and plankton that exist in lakes,

oceans, and rivers.

Figure 1.1. Natural light enters from air to an underwater


8/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

4

As light reflected from objects propagates toward the camera, a portion of the light meets

these suspended particles. This in turn absorbs and scatters the light beam, as illustrated in

Fig.1. In the absence of blackbody radiation, the multi-scattering process along the course of

propagation further disperses the beam into homogeneous background light.

Natural light enters from air to an underwater scene point

Light scattering and color change can be corrected by contrast enhancement and image

dehazing techniques. In this paper the performance of image enhancement using Dark

channel prior is compared with that of Histogram equalization.

Underwater vision is one of the scientific fields of

investigation for researchers. Autonomous Underwater Vehicles (AUV) and RemotelyOperated Vehicles (ROV) are usually employed to capture the data such as underwater

mines, shipwrecks, coral reefs, pipelines and telecommunication cables from the underwater

environment.


9/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

5

2.1HISTOGRAM EQUILIZATION

Histogram equalization (HE) is popular method of contrast adjustment using the images

histogram and also enhances a given image [3]. In this method, transformationT is to be

designed in such a way that the gray value in the output is equally distributed in [0, 1]. It is

also called histogram flattening. Histogram equalization method in which histogram is

modified by spreading the gray level areas. When an image's histogram is made equal, all

pixel values of the image are redistributed so there are approximately an equal number of

pixels to each of the user specified output gray-scale classes e.g., 32, 64, and 256. Contrast

is increased at the most populated range of brightness values of the histogram. For very

bright or dark parts of the image, it automatically reduces the contrast associated with the

ends of a normally distributed histogram. It can also divide pixels into different groups, if

few output values are over a wide range. But this method is effective only when the original

image has poor contrast to start with, otherwise it may degrade the image quality.

Fig 2.1: histogram equalization applied to low contrast image


10/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

6

2.2IMAGE DEHAZING

Dehazing is the process to remove haze effects in captured images and reconstruct the

original colors of natural scenes, which will be a useful pre-processing for input images and

required for receiving high performance of the vision algorithm. However, haze removal is a

challenging problem since the degradation is spatial-variant, it depends on the unknown

scene depth. In the literature, a few approaches have been proposed by using multiple

images or additional information. For example, polarization-based methods remove the haze

effect through two or more images taken with different degrees of polarization; scene

depths are estimated from multiple images of the same scene that are captured in different

weather conditions when using depth-based methods . Although these methods may produce

impressive results, they are impractical, because the requirements cannot always be satisfied

and make them difficult to meet with real-time applications of images with changing scenes.

In order to overcome the drawback of multiple images dehazing algorithms, many

researchers focus on achieving haze removal results from a single degraded image. Tan

developed the contrast maximization technique for haze removal relied on the observations

that images with enhanced visibility have more contrast than images plagued by bad

weather. Under the assumption that the transmission and the surface shading are locally

uncorrelated, Fatal presented a method for estimating the transmission in hazy scenes. He

propose a novel priordark channel priorfor single image haze removal, which is based

on the statistics of outdoor haze-free images. Combining a haze imaging model and a soft

matting interpolation method, they can recover a high-quality haze-free image.

When one takes a picture in foggy weather

conditions, the obtained image often suffers from poor visibility. The distant objects in the

fog lose the contrasts and get blurred with their surroundings. This is because the reflectedlight from these objects, before it reaches the cameras attenuated in the air and further

blended with the atmospheric light scattered by some aerosols (e.g., dust and water-

droplets). Also for this reason, the colors of these objects gets faded and become much

similar to the fog, the similarity of which depending on the distances of them to the camera.

The image dehazing is essentially an under-constrained problem.

The general principle of solving such problems is therefore to explore additional priors or


11/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

7

constraints. Following this idea, deriving an inherent boundary constraint on the scene

transmission. This constraint combined with a weighted L1 norm based contextual

regularization between neighboring pixels, is formalized into an optimization problem to

recover the unknown transmission. Dehazing method requires only a few general

assumptions and can restore a haze-free image of high quality with faithful colors and fine

edge details.

Fig2.2: The original image with fog and the present DE hazed image result


12/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

8

2.3WCID Process

A WCID means wavelength compensation and image dehazing. WCID is an underwater

image enhancement method or technique which compensate wavelength. As discussed

above, two main causes of underwater image distortions are light scattering and color

change. Sometimes artificial light source is used to overcome insufficient lightening

problem. But it introduce additional luminance in the image. WCID is an only technique

which handles problems of light scattering, color change and artificial light source presence

simultaneously. This technique has a novel systematic approach to enhance underwater

images by dehazing algorithm. A number of underwater image processing techniques used

to remove light scattering and color change effect. Generally most of the processing

techniques focus on removing either light scattering effect or color change effect. The only

technique called WCID will handle these problems simultaneously.

Fig 2.3: image formation model

A WCID is based on an underwater image formation. the algorithm for wavelength

compensation and image dehazing combines techniques to remove distortion caused by

light scattering and color change .First ,Dark channel prior method is used to estimate

distance between camera and object .based on the depth map derived ,the foreground and

background areas are segmented .the light intensities of foreground and background are

compared to determine presence of artificial light .if artificial light source is detected ,the

luminance introduced by it is removed from the foreground area. Next dehazing algorithm is

used to remove haze effect and color change.


13/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

9

Fig 2.4: WCID flow diagram

Fig 5 shows WCID flow diagram. It shows how WCID algorithm will work. The WCID

algorithm combines techniques of WCID to remove distortions caused by light scattering

and color change. Dark channel prior, an existing scene depth method is used to estimate

distance between object and camera. Based on depth map derived, the foreground and

background areas within the image are segmented. The intensities of foreground and

background then compared to determine whether there is presence of artificial light. If

artificial light source is detected, then luminance which is contributed by artificial is

removed to avoid overcompensation. Then the dehazing algorithm and wavelength

compensation are utilized to remove color change effect. The residual energy among

different color channels is used to estimate water depth. Following steps are required in

WCID algorithm.

Step 1: calculation of distance between camera and object d(x)

A dark channel prior method is used to estimate d(x). The low intensities in the dark channel

are due to shadows, colorful objects and dark objects. Depth map required two images for

parallax. Generally haze increases as distance increase. The dark channel prior method is

based on the observation that in most of the non-background light patches; at least one color

channel has a very low intensity at some pixels.


14/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

10

Step 2: Removal of artificial light source

Sometimes artificial light sources are used in underwater photography for obtaining clear

image. This light source contributed additional luminance which must be deducted to avoid

overcompensation. For this, we compare foreground and background luminance and then

compare this luminances to remove luminance contributed by artificial light source.

Step 3: Calculation of underwater depth (D)

In WCID algorithm, underwater depth is calculated by comparing energy for channels. For

above water surface, energy is same for red, green and blue channels. But for below water

surface energy of color channels depends on their attenuation.

Step 4: calculation of image depth range (R)

When light penetrates through water covering the depth range R, a disparate amount of color

change will be induced at the top and at the bottom of the image. This phenomenon

necessitates varying energy compensation adjusted according to the depth of each

underwater point to correctly rectify color change.


15/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

11

3.DARK CHANNEL PRIOR

Dark channel prior method can produce a natural haze free image. However, because this

approach is based on a statistically independent assumption in a local patch, it requires the

independent components varying significantly. Any lack of variation or low signal-to-noise

ratio (e.g., in dense haze region) will make the statistics unreliable. Moreover, as the

statistics is based on color information, it is invalid for grayscale images and difficult to

handle dense haze which is often colorless and prone to noise [4]. Main contributions of

dark channel method as follows: 1) The Separating an image into diffuse and specular

components is an ill-posed problem due to lack of observations. 2) The observed color of an

image is formed from the spectral energy distributions of the light reflected by the surface

reflectance, and the intensity of the color is determined by the imaging geometry. 3) The

dark channel is taken from the lowest intensity value among RGB channels at each pixel.

The DCP approach is able to handle distant objects even in the heavy haze image.

It does not rely on significant variance on transmission or surface shading in the input

image. The result contains few halo artifacts. Like any approach using a strong assumption,

our approach also has its own limitation. The dark channel prior may be invalid when the

scene object is inherently similar to the air light over a large local region and no shadow is

cast on the object. The complete dehazing procedure is summarized in Algorithm 1 as

shown in fig 2.

Algorithm 1: Underwater Image Haze and Noise Removal via Dark Channel Followed by

Dehazing.

1. Estimate I by Color models contrast variation in input image Y

2. Estimate transmission, t, and atmospheric light (a) at infinity. (e.g. dark channel prior)

from I

3. Dehaze I through inversion of Eq. (1) to obtain the restored image: the scene radiance

estimate R (e.g. dark channel prior) from I.


16/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

12

A.Color Feature Extraction

A color image is a combination of some basic colors. In each individual pixel of

a color image (termed true color) down into Red, Green and Blue values. We are going to

get as a result, for the entire image is 3 matrices, each one representing color features. The

three matrices are arranging in sequential order, next to each other creating a 3 dimensionalm by n by 3 matrixes.

1) Three color planes namely Red, Green and Blue are separated.

2) for each plane row mean and column mean of colors are calculated.

3) The average of all row means and all columns means are calculated for each color plane.

4) The features of all 3 planes are combined to form a feature vector. Once the feature

vectors are generated for all images in the database, they are stored in a feature data set.

B.Estimating the Light Source of Illumination

The underwater image color constancy algorithms are based on simplifying assumptions

such as restricted (limited number of image colors which can be observed under a specific

illuminant), the distribution of colors that are present in an image (e.g. white patch) and the

set of possible light sources. The dimensions should be Px3, where P is the number of data

points and 3 is the number of color channels.

To estimate the color of the light source to compute the canonical ranges for the lightillumination type (i.e., pixels, Edges and Gradient). For accurate performance, the canonical

range must be learned using images that are a representative set of real-world surfaces. Also,

all images that are used to learn the canonical range must be images that are illuminated by

the same light source. To take illumination from all directions into account, let us consider

an infinitesimal patch of extended light source of an image.

The Color transmission algorithm is based on the average reflectance in a scene under a

neutral light source is achromatic. In these two algorithms are proven to be important

instantiations of the Murkowski-norm.

Lc (P) = ( Fcp (x) dx ) 1/p = KEc

Where c = {R, G, B} and k is a multiplicative constant chosen such that the illuminant color,

e = (eR, eG, eB) T.


17/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

13

C.Dark channel estimation

The dark channel prior is based on the following observation on haze-free outdoor images:

in most of the non-sky patches, at least one color channel has very low intensity at some

pixels. In other words, the minimum intensity in such a patch should have a very low value.

Formally, for an image I, we define

Where Ic is a color channel of I and (x) is a local patch centered at x. our observation

says that except for the water color region the intensity of I dark is low, and tends to be zero,

if I is the haze free underwater image.

The low intensities in the dark channel are mainly due to three factors:a) Shadows. e.g., the shadows of leaves, trees and rocks in landscape images.

b) Colorful objects or surfaces. e.g., any object (for example, green grass/tree/plant, red or

yellow flower/leaf, and blue water surface) lacking color in any color channel will result in

low values in the dark channel.

c) Dark objects or surfaces. e.g., dark tree trunk and stone. As the natural outdoor images

are usually full of shadows and colorful, the dark channels of these images are dark.


18/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

14

D.Haze removal

The dark channel prior relies on sample minima; it is especially sensitive to outliers [9].

Various approaches can be taken to robustly estimate the dark channel, and by extension the

transmission map, considering the presence of noise. The hazy image prior to performing

the dehazing process is a natural approach to handling the problem of noise in underwater

scene radiance recovery.

In our restoration scheme, dehazing as a pre-processing step is

especially convenient considering that it is already necessary for estimating the atmospheric

light and transmission map. In the dehazing step, we can treat our image model as: Y = I +

n, with the task being only to estimate I, which encapsulates the hazy image. After the hazyimage is dehazed, the rest of the dehazing process is exactly the same as in the noise-free

case.

Fig 3.1:Flow Chart of Dehazing Process


19/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

15

4. EXPERIMENTAL RESULTS

The performance of the dark channel algorithm is evaluated for underwater images. and

videos downloaded from youtube. Results demonstrate haze removing of the dark channel

prior algorithm over histogram equalization. Fig.3 and Fig.4 shows the result after

processing with a dark channel prior algorithm and histogram equalization. Performance

evaluation of both dark channel prior and histogram equalization is shown in Table 1.

fig4.1: (A) underwater image-enhanced image using (b) dark prior channel

(c) Histogram equalization


20/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

16

Comparison between the dark prior channel and the histogram equalization is shown in the

table below. This table shows the quantitative performance analysis of dark prior channel

over histogram equalization.

TABLE4.1: QUANTITATIVE PERFORMANCE ANALYSIS

PSNR

Underwater image1 Underwater

image2

Dark Channel Prior 50.4035dB 48.7065dB

Histogram Equalization 25.3084dB 22.2872dB


21/22


DEPT.OF ECE CUCEK

SEPTEMBER 2015

17

CONCLUSION

The Dehazing algorithm effectively restore image color balance and remove haze over

histogram equalization. No existing techniques can handle light scattering and color change

distortions suffered by underwater images simultaneously. The experimental results

demonstrate superior haze removing and color balancing capabilities of the DCP algorithm

over traditional Dehazing and color methods. However, the salinity and the amount of

suspended particles in ocean water vary with time, location, and season, making accurate

measurement of the rate of light energy loss is difficult. Errors in the rate of light energy loss

will affect the precision of both the water depth and the underwater propagation distance. A

very simple but powerful prior is called dark channel prior, for underwater image hazeremoval. The dark channel prior is based on the statistics of the underwater images.

Applying the prior into the haze imaging model, haze can be effectively removed. Though,

haze can be removed effectively in this paper, color change distortion still exist in the

underwater image. Wavelength compensation of enhancement technique to be used for color

change distortion in future.


22/22

Underwater Image Enhancement SEPTEMBER 2015

REFERENCES

[1]. K. Lebart, C. Smith, E. Trucco, and D. M. Lane, Automatic indexing of underwater

survey video: algorithm and benchmarking method, IEEE J. Ocean. Eng., vol. 28, no. 4, pp.

673686, Oct. 2003.

[2]. Prabhakar C.J.1*, Praveen Kumar P.U, An Image Based Technique for Enhancement

of Underwater mages , December 09, 2011

[3]. J. R. Zaneveld and W. Pegau, Robust underwater visibility parameter, Opt. Exp., vol.

11, no. 23, pp. 29973009, 2003.

[4]. E. Trucco and A. T. Olmos-Antillon, Self-tuning underwater image restoration, IEEE

J. Ocean. Eng., vol. 31, no. 2, pp. 511519, Apr. 2006.

[5]. J. S. Jaffe, Computer modeling and the design of optimal underwater imaging

systems, IEEE J. Ocean. Eng., vol. 15, no. 2, pp. 101111, Apr. 1990.

[6]. M. C.W. van Rossum and T. M. Nieuwenhuizen, Multiple scattering of classical

waves: Microscopy, mesoscopy and diffusion, Rev. Modern Phys., vol. 71, no. 1, pp. 313

371, Jan. 1999.

[7]. Y. Y. Schechner and N. Karpel, Recovery of underwater visibility and structure by

polarization analysis, IEEE J. Ocean. Eng., vol. 30, no. 3, pp. 570587, Jul. 2005.

[8]. I. Vasilescu, C. Detwiler, and D. Rus, Color-accurate underwater imaging using

perceptual adaptive illumination, in Proc. Robot. Sci. Syst., Zaragoza, Spain, 2010.

[9]. K. He, J. Sun, and X. Tang, Single image haze removal using Dark Channel Prior, in

Proc.IEEE CVPR, 2009, vol. 1, pp. 19561963.

Seminar Report(Chiranjivi)

Documents

Transcript of Seminar Report(Chiranjivi)