Noise Minimization in Images and Communication

MAJOR PROJECT
REPORT
ODD SEM (2012-2013)

2012
ALOK SHIVAM (9909102226) and G.KRISHNA CHAITANYA(9909102223)
B.TECH(ECE), JAYPEE INSTITUTE OF INFORMATION TECHNOLOGY
12/12/2012
MAJOR PROJECT REPORT- NOISE MINIMIZATION TECHNIQUES
2

CERTIFICATE

This is to certify that the work titled NOISE MINIMIZATION
TECHNIQUES submitted by ALOK SHIVAM and G. KRISHNA
CHAITANYA in partial fulfillment for the award of degree of
bachelors of technology (`ECE) of Jaypee Institute of Information
Technology University, Noida has been carried out under my
supervision. This work has not been submitted partially or wholly to
any other University or Institute for the award of this or any other
degree or diploma.

Signature of Supervisor ..
Name of Supervisor Mr. KAPIL DEV TYAGI
Designation Sr. LECTURER
Date 12
TH
DECEMBER, 2012

3

ACKNOWLEDGEMENT

We would like to take this opportunity to express our gratitude towards our
teacher, project co-coordinator and mentor KAPIL DEV TYAGI SIR for his
undying support and fruitful directives to help us in making this project, and
boosting our confidence. We salute his able guidance and suggestions
without which the project would have been almost impossible to meet its
goal. We take immense pleasure in thanking JIIT, NOIDA for providing us
with an opportunity to work at the temple of technology and providing us an
opportunity to work in their PROJECT LAB. We would like to give special
thanks to Mr. B. Suresh sir for being a constant source of guidance
throughout the project.
We would also like to thank our parents who boosted us morally and were a
constant support to us and acted like a backbone for the project. We would
like to thank god for all the blessings he bestowed upon us and for giving us
the strength to do this project.

ALOK SHIVAM

G.KRISHNA CHAITANYA

4

OBJECTIVES OF THE PROJECT

1. TO STUDY AND IMPLEMENTAION OF NOISE
MINIMIZATION TECHNIQUES
a. In images
b. In communication system
c. In wireless communication

2. To study algorithms that could reduce noise during signal
transmission.

3. To study research works and research papers to develop an
algorithm on own

4. To analyze the effect of noise from different transmission
angles.

5. To carry out real time noise minimization

6. To work for noise minimization in video signals.

5

CHAPTER 1:- INTRODUCTION

With communication advancing day by day, noise reduction is one area which has been a
concern for communication during all times. Todays advancements have come up with
highly efficient tools and techniques for noise minimization.
Noise is one such signal which is not in the control of the provider or the user, it can be in
any form and providing a noise free communication to the user is a big challenge for
modern day industries.
In common use, the word noise means any unwanted sound. In physics and analogue
electronics, noise is a mostly unwanted random addition to a signal; it is called noise as a
generalisation of the acoustic noise, heard when listening to a weak radio transmission
with significant electrical noise. Signal noise is heard as acoustic noise if the signal is
converted into sound (e.g., played through a loudspeaker). High noise levels can block,
distort, change or interfere with the meaning of a message in human, animal and
electronic communication. In signal processing or computing it can be considered
random unwanted data without meaning; that is, data that is not being used to transmit a
signal, but is simply produced as an unwanted by-product of other activities. "Signal-to-
noise ratio is sometimes used to refer to the ratio of useful to irrelevant information in
an exchange.
Noise reduction is the process of removing noise from a signal. All recording devices,
both analogue and digital, have traits which make them susceptible to noise. Noise can be
random or white noise with no coherence, or coherent noise introduced by the device's
mechanism or processing algorithms.
In electronic recording devices, a major form of noise is hiss caused by random electrons
that, heavily influenced by heat, stray from their designated path. These stray electrons
influence the voltage of the output signal and thus create detectable noise.
This report presents various aspects of noise in images, wired and wireless
communication, study and implementation of minimization algorithms on matlab,
comparisons between the techniques and providing results and analysis of the research
work done. Finally the most effective and simple technique is recommended for
implementation for each type of communication.

6

CHAPTER 2:- NOISE MINIMIZATION IN IMAGES

2.1 INTRODUCTION
Images taken with both digital cameras and conventional film cameras will pick
up noise from a variety of sources. Further uses of these images require that the
noise will be (partially) removed for artistic work or marketing, or for practical
purposes such as computer vision.
. 2.1.1 Types of noise in images
In salt and pepper noise (sparse light and dark disturbances), pixels in the image
are very different in color or intensity from their surrounding pixels; the defining
characteristic is that the value of a noisy pixel bears no relation to the color of
surrounding pixels. Generally this type of noise will only affect a small number of
image pixels. When viewed, the image contains dark and white dots, hence the
term salt and pepper noise. Typical sources include flecks of dust inside the
camera and overheated.
In Gaussian noise, each pixel in the image will be changed from its original value
by a (usually) small amount. A histogram, a plot of the amount of distortion of a
pixel value against the frequency with which it occurs, shows a normal
distribution of noise. While other distributions are possible, the Gaussian (normal)
distribution is usually a good model, due to the central limit theorem that says that
the sum of different noises tends to approach a Gaussian distribution.
In either case, the noise at different pixels can be either correlated or uncorrelated;
in many cases, noise values at different pixels are modeled as being independent
and identically distributed, and hence uncorrelated.
In selecting a noise reduction algorithm, one must weigh several factors:

- whether sacrificing some real detail is acceptable if it allows more noise to be
removed (how aggressively to decide whether variations in the image are noise or
not)
- the characteristics of the noise and the detail in the image, to better make those
decisions
Digital cameras have become very popular over the last several years for both
professional and personal use. Under most conditions, digital cameras are an efficient,
convenient way to take pictures.The development of a noise reduction technique which
can be used to decrease the visible noise in images taken using exposure times greater
than 1/4 second.Using this noise reduction technique, people who have taken images in
low-light situations are able to take advantage of the many benefits of digital
photography.Images taken with both digital cameras and conventional film cameras will
pick up noise from a variety of sources. further uses of these images require that the noise
will be (partially) removed - for aesthetic purposes as in artistic work or marketing, or for
practical purposes such as computer vision.
7

2.2 NOISE REDUCTION METHODS :

2.2.1.Pixel Substitution :
The first method utilized to reduce the noise consisted of substituting one pixel value for
another. A program was written to scan through a dark image looking for pixels with
values above the threshold input to the program. These pixels will be referred to as 'hot'
pixels. When a hot pixel was found, the program located the lowest pixel value within
two pixels from the hot pixel. The locations of each of the hot and cold pixels were
recorded in a data file used when processing images. For each of the hot pixel locations,
the digital count in that location was replaced by the value in the corresponding cold
pixel.
Figures 5 and 6 show the results of the substitution method. A reduction in the amount of
noise is evident in the processed image.

Figure 1: Original 30 Second Image Figure 2: Pixel Substitution

Additionally, an artifact was discovered at many of the edges in the image. The artifat
was caused by having a hot pixel in the dark part of the image and the corresponding cold
pixel in the lighter area.
While some reduction in the noise has been achieved, the presence of this edge artifact
was unacceptable, so another approach had to found.

8

2.2.2 SALT AND PEPPER :
In salt and pepper noise (sparse light and dark disturbances), pixels in the image are very
different in color or intensity from their surrounding pixels; The defining characteristic is
that the value of a noisy pixel bears no relation to the color of surrounding pixels.
Generally this type of noise will only affect a small number of image pixels. When
viewed, the image contains dark and white dots, hence the term salt and pepper noise.
Typical sources include flecks of dust inside the camera and overheated.This type of
noise can be caused by dead pixels, analog-to-digital converter errors, bit errors in
transmission, etc.This can be eliminated in large part by using dark frame subtraction and
by interpolating around dark/bright pixels.

CODE:
WITH USING FUNCTIONS:
I = imread('C:\Users\9909102213\Desktop\nayanatara.gif');
imshow(I)
J = imnoise(I,'salt & pepper',0.02);
figure, imshow(J)
K = filter2(fspecial('average',3),J)/255;
figure, imshow(K)
L = medfilt2(J,[3 3]);
figure, imshow(L)

WITH OUT USING MATLAB FUNCTIONS:
A = imread('C:\Users\9909102213\Desktop\nayanatara.gif');
imshow(A)
J = imnoise(A,'salt & pepper',0.02);
figure, imshow(J)
%PAD THE MATRIX WITH ZEROS ON ALL SIDES
modifyA=zeros(size(J)+2);
B=zeros(size(J));
for x=1:size(J,1)
for y=1:size(J,2)
modifyJ(x+1,y+1)=J(x,y);
end
end
9

for i= 1:size(modifyJ,1)-2
for j=1:size(modifyJ,2)-2
window=zeros(9);
inc=1;
for x=1:3
for y=1:3
window(inc)=modifyJ(i+x-1,j+y-1);
inc=inc+1;
end
end
med=sort(window);
B(i,j)=med(5);
end
end
B=uint8(B);
figure,imshow(B);

10

Fig 3 fig 4

2.2.3 GUASSIAN NOISE
In Gaussian noise, each pixel in the image will be changed from its original value by a
(usually) small amount. A histogram, a plot of the amount of distortion of a pixel value
against the frequency with which it occurs, shows a normal distribution of noise. While
other distributions are possible, the Gaussian (normal) distribution is usually a good
model, due to the central limit theorem that says that the sum of different noises tends to
approach a Gaussian distribution.
In either case, the noise at different pixels can be either correlated or uncorrelated; in
many cases, noise values at different pixels are modeled as being independent and
identically distributed, and hence uncorrelated.

CODE:
RGB = imread('C:\Users\9909102213\Desktop\nayanatara.gif');
I = rgb2gray(RGB);
J = imnoise(I,'gaussian',0,0.025);
imshow(J)
L = filter2(fspecial('average',3),J)/255;
figure, imshow(L)
K = wiener2(J,[5 5]);
figure, imshow(K);

11

12

2.2.4 POISSION NOISE :
Poisson noise or shot noise is a type of electronic noise that occurs when the finite
number of particles that carry energy, such as electrons in an electronic circuit or photons
in an optical
device, is small enough to give rise to detectable statistical fluctuations in a measurement.
2.2.4.1 CODE:

I = rgb2gray(RGB);
J = imnoise(I,'poisson');
imshow(J) ;
figure, imshow(L);
K = wiener2(J,[5 5]);
figure, imshow(K) ;

13

2.2.5 SPECKLE NOISE:
Speckle noise is a granular noise that inherently exists in and degrades the quality of the
active radar and synthetic aperture radar (SAR) images. Speckle noise in conventional
radar results
from random fluctuations in the return signal from an object that is no bigger than a
single image-processing element. It increases the mean grey level of a local area. Speckle
noise in SAR is generally more serious, causing difficulties for image interpretation. It is
caused by coherent processing of backscattered signals from multiple distributed targets.
In SAR oceanography, for example, speckle noise is caused by signals from elementary
scatters, the gravity-capillary ripples, and manifests as a pedestal image, beneath the
image of the sea waves.

2.2.5.1 CODE:
I = rgb2gray(RGB);
J = imnoise(I,'speckle',v)
imshow(J)
figure, imshow(L)
K = wiener2(J,[5 5]);
figure, imshow(K)

14

2.2.5.2 CODE:- COINS
A=imread('C:\Users\9909102213\Desktop\coins.gif');

B=bitget(A,1);
figure,
subplot(2,2,1);imshow(logical(B));title('Bit plane 1');

B=bitget(A,2);

B=bitget(A,3);

B=bitget(A,4);
15


B=bitget(A,5);
figure,

B=bitget(A,6);

B=bitget(A,7);

B=bitget(A,8);

RECONSTRUCTION:-
A=imread('C:\Users\9909102213\Desktop\coins.gif');
B=zeros(size(A));
B=bitset(B,8,bitget(A,8));

B=uint8(B);
figure,imshow(B);

2.2.6 CONCLUSION:
We used the Nayantara Image(figure 1) in png format ,adding four noise (Speckle,
Gaussian ,Poisson and Salt & Pepper) in original image with standard deviation(0.025)
De-noised all noisy images by all filters and conclude from the results that:
(a)The performance of the Wiener Filter after de-noising for all Speckle, Poisson and
Gaussian noise is better than Mean filter and Median filter.
(b)The performance of the Median filter after de-noising for all Salt & Pepper noise is
better than Mean filter and Wiener filter.

16

CHAPTER 3:- NOISE MINIMIZATION IN COMMUNICATION

3.1 BASICS OF COMMUNICATION
A COMMUNICATION SYSTEM:-
A communication system is set of individual communication networks, transmission
systems, relay stations, tributary stations, and data terminal equipments (DTE) usually capable of
interconnection and interoperation to form an integrated whole. The components of a
communications system serve a common purpose, are technically compatible, use common
procedures, respond to controls, and operate in unison. A modern day telephony system is an
example of communication system.

3.1.1 Sampling
Sampling is a process where you convert continuous time signals(analog signals)
into discrete time signals(digital system). On paper CT signals are represented by
"full lines" where as DT signals are only "points" chosen from the original CT
signals. Now it is clear that DT signals are a rough representation of their
respective CT signals. Now the question arises, why do we need sampling? We
need it because it is very efficient to use "rough" DT signals in
telecommunication instead of "precision" CT signals. Now you may ask, if we're
use "rough" DT signals doesn't the information that is to be transmitted and
henceforth received is also "rough" or inaccurate? That's where the sampling
theorem comes in.The sampling theorem gives you a rule using which you can
use DT signals to transmit/receive the information accurately. ST simply states
17

that the 'sampling frequency' should be greater than or equal to twice the
frequency of the CT signal, where sampling frequency is frequency of sampled
signals(DT signals) to be obtained by Sampling. It simply means that when you're
taking samples(choosing those "points" from "full line") from the CT signal, do it
in such a manner that the samples("chosen points") are more closely spaced. And
also take higher number of samples(that is choose higher number of points from
the CT signal).

3.1.2 Quantization
Quantization is the process of mapping a large set of input values to a smaller set such
as rounding values to some unit of precision. A device or algorithmic functions that
perform quantization is called a quantizer. The error introduced by quantization is
referred to as quantization error or round off error. Quantization is involved to some
degree in nearly all digital signal processing, as the process of representing a signal in
digital form ordinarily involves rounding. Quantization also forms the core of essentially all
lossy compensation algorithms.
Because quantization is a many-to-few mapping, it is an inherently non-linear and
irreversible process (i.e., because the same output value is shared by multiple input
values, it is impossible in general to recover the exact input value when given only the
output value).

3.1.3 Encoding
Encoding is the process by which information from a source is converted into
symbols to be communicated. Decoding is the reverse process, converting these
code symbols back into information understandable by a receiver.
One reason for coding is to enable communication in places where ordinary
spoken or written language is difficult or impossible. For example, semaphore,
where the configuration of flags held by a signaller or the arms of a semaphore
tower encodes parts of the message, typically individual letters and numbers.
Another person standing a great distance away can interpret the flags and
reproduce the words sent.

18

3.2 NOISE IN COMMUNICATION
Noise affects communication by somehow altering the message. In a very basic model of
communication, a message has to pass from the receiver to the recipient through a
channel. Noise is something that interferes with the transmission of the message through
the channel. An easy way to think of this is to picture static on a TV. The static is the
noise that interferes with the transmission of the TV program through the channel of your
cable box or satellite dish. This makes it difficult to understand the message sent by the
TV program; you may not hear what someone on TV says correctly, and so misinterpret
what they say, for example. Noise doesn't necessarily have to be something external that
affects the channel. It can also be something internal to the recipient that affects how they
receive the message. An example of this is if someone is in a bad mood. Being in a bad
mood makes you more likely to interpret other people's messages as negative or hostile;
the noise of your emotion affects how you see what they're trying to communicate.
3.2.1 Types of noise in communication are classified as:-

EXTERNAL NOISE: - noise external to the system is external noise
ATMOSPHERIC NOISE: - noise due to natural electric disturbances such as lightning.
SOLAR NOISE: - noise due to solar radiations
Industrial noise: - electric motor, generator, aircraft produce EM sig. In the range 1 Hz-
600 Hz.
INTERNAL NOISE: - noise internal to a system is internal noise
SHORT NOISE: - it arises in electric devices because of discrete nature of current flow,
it is Gaussian in nature.
FLICKER NOISE: - it is due to imperfect surface behaviour of semi-conductor devices
THERMAL NOISE: - NOISE generated due to random motion of free charged particles.
19

3.3 FILTERING
A filter is a device or process that removes from a signal some unwanted component or
feature. Filtering is a class of signal processing, the defining feature of filters being the
complete or partial suppression of some aspect of the signal. Most often, this means
removing some frequencies and not others in order to suppress interfering signals and
reduce background noise.
A low-pass filter is an electronic filter that passes low-frequency
signals and attenuates (reduces the amplitude of) signals with frequencies higher than
the cutoff frequency. The actual amount of attenuation for each frequency varies from
filter to filter. It is sometimes called a high-cut filter, or treble cut filter when used in
audio applications. A low-pass filter is the opposite of a high-pass filter. A band-pass
filter is a combination of a low-pass and a high-pass.
3.3.1 Butterworth filter
The Butterworth filter is a type of signal processing filter designed to have as flat
a frequency response as possible in the passband. It is also referred to as a maximally
flat magnitude filter.

20

3.3.2 Chebychev filter
Chebyshev filters are analog or digital filters having a steeper roll-off and
more pass band ripple (type I) or stop band ripple (type II) than Butterworth
filters. Chebyshev filters have the property that they minimize the error between
the idealized and the actual filter characteristic over the range of the filter, but
with ripples in the pass band.
Response of a 1
st
Oder chebychev filter.
3.3.3 Comparison of the response of each filters

21

3.3.4.2 Matlab code for designing a filter(5
th
Oder Butterworth filter)
fs = 1e4;
t = 0:1/fs:5;
n = sin(2*pi*1000000*t);
sw = 0.1*cos(2*pi*5*t);
swn = sw + n;
figure(1)
plot(sw);
figure(2)
plot(n);
figure(3)
plot(swn);

Fc = 500 ; %cut off frequency

w = Fc/(2*fs);
[b,a]=butter(5,w,'low'); %5th order butterworth LPF
[h,w]=freqz(b,a,1024);

y=filter(b,a,swn);
figure(4)
plot(y);

figure(5)
plotHandle=plot(w/pi*2*fs,abs(h));
set(plotHandle,'LineWidth',2.5);
title('Frequency Response of a 5th Order Butterworth LPF');
xlabel('Frequency (Hz)')
ylabel('Magnitude');
grid;

OUTPUT:-

a b
c d
Input signal, noise signal , received signal and the response of the filter respectively.
22

3.3.4.2 Matlab code for noise reduction using low pass filtering for recorded sound
[y,fs]=wavread('11863__medialint__nord-analog-howling-wind-storm
(1).wav');

ch1=y(:,1);
time=(1/44100)*length(ch1);
t=linspace(0,time,length(ch1));
figure(1)
plot(y);

L=length(ch1);
NFFT = 2^nextpow2(L); % Next power of 2 from length of y
Y = fft(y,NFFT)/L;
Y1=log10(Y);
figure(2)

f = fs/2*linspace(0,1,NFFT/2+1);
plot(f,2*abs(Y1(1:NFFT/2+1))) ;

[b,a]=butter(10,3000/(44100/2),'high');
Y1=filtfilt(b,a,Y1);

% freqz(b,a)
figure(3)

plot(f,2*abs(Y1(1:NFFT/2+1))) ;

title('Single-Sided Amplitude Spectrum of y(t)');
xlabel('Frequency (Hz)');
ylabel('|Y(f)|')
xlim([0 50000])
OUTPUT:-

23

3.4 EMPHERICAL MODE DECOMPOSITION ALGORITHM

4.1 INTRODUCTION:-

EMD is a method of breaking down a signal without leaving the time domain. It can be
compared to other analysis methods like Fourier Transforms and wavelet decomposition.
The process is useful for analyzing natural signals, which are most often non-linear and
non-stationary. This parts from the assumptions of the methods we have thus far learned
(namely that the systems in question be LTI, at least in approximation).

EMD is a method of breaking down a signal without leaving the time domain. It can be
compared to other analysis methods like Fourier Transforms and wavelet decomposition.
The process is useful for analyzing natural signals, which are most often non-linear and
non-stationary. This parts from the assumptions of the methods we have thus far learned
(namely that the systems in question be LTI, at least in approximation).
EMD filters out functions which form a complete and nearly orthogonal basis for the
original signal. Completeness is based on the method of the EMD; the way it is
decomposed implies completeness. The functions, known as Intrinsic Mode Functions
(IMFs), are therefore sufficient to describe the signal, even though they are not
necessarily orthogonal. The reasons are described in Huang et al., published in the Royal
Society Proceedings on Math, Physical, and Engineering Sciences: "...the real meaning
here applies only locally. For some special data, the neighbouring components could
certainly have sections of data carrying the same frequency at different time durations.
But locally, any two components should be orthogonal for all practical purposes" .
The fact that the functions into which a signal is decomposed are all in the time-domain
and of the same length as the original signal allows for varying frequency in time to be
preserved. Obtaining IMFs from real world signals is important because natural processes
often have multiple causes, and each of these causes may happen at specific time
intervals. This type of data is evident in an EMD analysis, but quite hidden in the Fourier
domain or in wavelet coefficients.
Some examples of data to which the EMD method may be applied quite effectively are
seismic readings, results of neuroscience experiments, electrocardiograms (which we will
examine later), gastroelectrograms, and sea-surface height (SSH) readings.
IMF:
1.The EMD will break down a signal into its component IMFs.
2.An IMF is a function that:
- has only one extreme between zero crossings, and
- has a mean value of zero.

24

4.2 APPLICATION:
EMD is most useful (and as we will see later, perhaps only useful) for non-linear, non-
stationary signals, and natural signals. As an example of this, we have applied the EMD
to several signals, two of which are the raw data from electrocardiograms that we found
on the web. The horizontal axis represents sample number, as sampling rate was not
available for both.

Fig4.2.1 ECG signal

Fig4.4.2 Plots run from c
14
on the upper left to c
1
on the lower right.

25

Each IMF represents a different part of the signal, giving a fairly good breakdown of
different causation parts of the total composite heartbeat. While the wave in ecg.mat (of
which we still cannot determine a cause) causes the EMD some problems, the signal
ekg.mat was broken down fairly efficiently, especially in IMF c
3
, where each heartbeat
has been identified as a single entity amidst the other lesser parts that together compose a
single beat.

Part of the importance of the EMD is the way it decomposes signals over which Fourier
Transforms stumble. Because of the way the EMD keeps signals in their own domain, it
can deal with signals which would otherwise be considered "poorly-behaved." When a
signal enters a new domain, the way certain select characteristics change with the original
variable are totally lost. While the signal may be accurately retrieved, the signal cannot
be effectively analyzed in the new domain without this information.
The EMD does not have this problem. When it decomposes a time-domain signal into
IMFs, for instance, each mode function contains information about how the frequency of
the original signal changes in time. Because of this, the EMD needs not assume the
slightest pretention towards linearity or time-invariance.

(a) the blue curve is the input signal x(t), red circles represent the local maxima, and the green squares
are local minima. (b) Black line is upper and lower envelopes represented by cubic spline interpolation,
and the red line is the mean envelope m11(t). (c) the blue curve represents the input signal minus the
mean envelop (h11 (t) = x(t) m11 (t)), and the black line is the envelopes. (d) The blue signal is the
first IMF (c1 (t)) since it meets the IMF requirements. (e) Blue curve is the input signal minus the first
IMF (residual r1(t) = x(t) c1 (t)), to be considered as new input signal. (f) Blue curve is the second IMF
(c2 (t)) together with its upper, lower and mean envelopes
26

4.3 CODE:

CODE 1 :
% EMD: Emprical mode decomposition
x = ecg(500)';
%x1 = 3.5*ecg(2700).';
%y1 = sgolayfilt(kron(ones(1,13),x1),0,21);
%n = 1:30000;
%del = round(2700*rand(1));
%mhb = y1(n + del);
%t = 0.00025:0.00025:7.5;
%plot(t,mhb);
%axis([0 2.5 -4 4]);
%grid;
%xlabel('Time [sec]');
%ylabel('Voltage [mV]');
%title('Maternal Heartbeat Signal');
%t=-10:0.1:10;
%x=2*sin(2*pi*t);
%n=10;
imf = emd(x,n)
%
% x - input signal (must be a column vector)
% n - number of intrinsic mode functions
% imf - Matrix of intrinsic mode functions (each as a row)
subplot(4,1,1);
plot(t,x)
subplot(4,1,2);
plot(t,x1)
subplot(4,1,3);
plot(t,x2)
subplot(4,1,4);
plot(t,x3)
27

CODE 2:

function imf = emd(x,n);
c = x(:)'; % copy of the input signal (as a row vector)
N = length(x);
%-------------------------------------------------------------------------
% loop to decompose the input signal into n successive IMFs
imf = []; % Matrix which will contain the successive IMF, and the residue
for t=1:n % loop on successive IMFs
%-------------------------------------------------------------------------
% inner loop to find each imf
h = c; % at the beginning of the sifting process, h is the signal
SD = 1; % Standard deviation which will be used to stop the sifting process
while SD > 0.3 % while the standard deviation is higher than 0.3 (typical value)
% find local max/min points
d = diff(h); % approximate derivative
maxmin = []; % to store the optima (min and max without distinction so far)
for i=1:N-2
if d(i)==0 % we are on a zero
if sign(d(i-1))~=sign(d(i+1)) % it is a maximum
maxmin = [maxmin, i];
end
elseif sign(d(i))~=sign(d(i+1)) % we are straddling a zero so
maxmin = [maxmin, i+1]; % define zero as at i+1 (not i)
end
end
if size(maxmin,2) < 2 % then it is the residue
break
end
% divide maxmin into maxes and mins
if maxmin(1)>maxmin(2) % first one is a max not a min
28

maxes = maxmin(1:2:length(maxmin));
mins = maxmin(2:2:length(maxmin));
else % is the other way around
maxes = maxmin(2:2:length(maxmin));
mins = maxmin(1:2:length(maxmin));
end
% make endpoints both maxes and mins
maxes = [1 maxes N];
mins = [1 mins N];

%-------------------------------------------------------------------------
% spline interpolate to get max and min envelopes; form imf
maxenv = spline(maxes,h(maxes),1:N);
minenv = spline(mins, h(mins),1:N);
m = (maxenv + minenv)/2; % mean of max and min enveloppes
prevh = h; % copy of the previous value of h before modifying it
h = h - m; % substract mean to h
% calculate standard deviation
eps = 0.0000001; % to avoid zero values
SD = sum ( ((prevh - h).^2) ./ (prevh.^2 + eps) );
end
imf = [imf; h]; % store the extracted IMF in the matrix imf
% if size(maxmin,2)<2, then h is the residue
% stop criterion of the algo. if we reach the end before n
if size(maxmin,2) < 2
break
end
c = c - h; % substract the extracted IMF from the signal
end
return

29

3.5 Least mean square algorithm
3.5.1 Stages of LMS filter
Stage I (ALE Filter)
ALE is used to cancel out sinusoidal noise from the speech signal. The figure shown
below elaborates the basic working of ALE filter.

The input of this filter is a delayed version of the primary signal. The delay is set in
such a way so as to de-correlate wideband noise in order to cancel out the sinusoidal
noise from the primary signal.
Stage II (NLMS Filter)
The Normalized Least Mean Square (NLMS) algorithm is a modified form of Least
Mean Square (LMS) algorithm, the NLMS filter is used to remove the wideband
noise using the signals obtained at the output of the ALE stages. The NLMS filter as
an adaptive noise canceller is shown in figure

30

3.5.2 Denoising using LMS algorithm
Least mean squares (LMS) algorithms are a class of adaptive filter used to mimic a
desired filter by finding the filter coefficients that relate to producing the least mean
squares of the error signal (difference between the desired and the actual signal)
The basic idea behind LMS filter is to approach the optimum filter weights , by
updating the filter weights in a manner to converge to the optimum filter weight. The
algorithm starts by assuming a small weights (zero in most cases), and at each step,
by finding the gradient of the mean square error, the weights are updated. That is, if
the MSE-gradient is positive, it implies, the error would keep increasing positively, if
the same weight is used for further iterations, which means we need to reduce the
weights. In the same way, if the gradient is negative, we need to increase the weights.
So, the basic weight update equation is :

where represents the mean-square error. The negative sign indicates that,
we need to change the weights in a direction opposite to that of the gradient slope.
The mean-square error, as a function of filter weights is a quadratic function
which means it has only one extrema, that minimises the mean-square error, which is
the optimal weight. The LMS thus, approaches towards this optimal weights by
ascending/descending down the mean-square-error vs filter weight curve.
drawback of the "pure" LMS algorithm is that it is sensitive to the scaling of its
input.
The Normalised least mean squares filter (NLMS) is a variant of the LMS
algorithm that solves this problem by normalising with the power of the input and
selecting optimal learning rate

31

3.5.3 Matlab code for implementation of LMS algorithm for noise minimization
close all;
clear all;clc;

t=1:0.025:5;
desired=5*sin(2*3.*t);

noise=5*sin(2*50*3.*t);

refer=5*sin(2*50*3.*t+ 3/20);

primary=desired+noise;

subplot(4,1,1);
plot(t,desired);
ylabel('desired');

subplot(4,1,2);
plot(t,refer);
ylabel('refer');

subplot(4,1,3);
plot(t,primary);
ylabel('primary');

order=2;
mu=0.005;
n=length(primary)
delayed=zeros(1,order);
adap=zeros(1,order);
cancelled=zeros(1,n);

for k=1:n,
delayed(1)=refer(k);
y=delayed*adap';
cancelled(k)=primary(k)-y;
adap = adap + 2*mu*cancelled(k) .* delayed;
delayed(2:order)=delayed(1:order-1);
end

subplot(4,1,4);
plot(t,cancelled);
ylabel('cancelled');

32

3.5.4 Output and Results

33

CHAPTER 4:- NOISE MINIMIZATION IN WIRELESS COMMUNICATION
4.1 What is LTE??
4.1.1 Introduction
LTE an initialise of Long Term Evolution, marketed as 4G LTE, is a standard
for wireless communication of high-speed data for mobile phones and data
terminals. It is based on the GSM/EDGE and UMTS/HSPA network technologies,
increasing the capacity and speed using a different radio interface together with
core network improvements. The standard is developed by the 3GPP (3rd
Generation Partnership Project) and is specified in its Release 8 document series,
with minor enhancements described in Release 9.
The world's first publicly available LTE service was launched
by TeliaSonera in Oslo and Stockholm on 14 December 2009. LTE is the natural
upgrade path for carriers with GSM/UMTS networks, but even CDMA holdouts
such as Verizo Wireless, who launched the first large-scale LTE network in North
America in 2010, and au by KDDI in Japan have announced they will migrate to
LTE. LTE is, therefore, anticipated to become the first truly global mobile phone
standard, although the use of different frequency bands in different countries will
mean that only multi-band phones will be able to use LTE in all countries where it
is supported.
LTE aims for:-
1. Higher performance
2. Backwards compatible
3. Wide application
4.1.2 Basics of LTE

OFDMA-orthogonal frequency division multiple access
SC-FDM- Single carrier FDM
34

4.2 OFDM in LTE
4.2.1 OFDM system based on FFT

4.2.1.1 Serial to parallel convertor and concepts.
In a conventional serial data system, the symbols are transmitted
sequentially, with the frequency spectrum of each data symbol allowed to
occupy the entire available bandwidth.
When the data rate is sufficient high, several adjacent symbols may be
completely distorted over frequency selective fading or multipath delay
spread channel.
An individual data element normally occupies only a small part of
available bandwidth.
Because of dividing an entire channel bandwidth into many narrow sub
bands, the frequency response over each individual sub channel is
relatively flat.
A parallel data transmission system offers possibilities for alleviating this
problem encountered with serial systems.
Resistance to frequency selective fading

35

4.2.1.2 Modulation/Mapping
The process of mapping the information bits onto the signal constellation plays a
fundamental role in determining the properties of the modulation.
An OFDM signal consists of a sum of sub-carriers, each of which contains M-ary
phase shift keyed (PSK) or quadrature amplitude modulated (QAM) signals.
Modulation types over OFDM systems
Phase shift keying (PSK)
Quadrature amplitude modulation (QAM)
M-ary phase shift keying
Consider M-ary phase-shift keying (M-PSK) for which the signal set is

where is the signal energy per symbol, is the symbol duration, and is
the carrier frequency.
This phase of the carrier takes on one of the M
possible values, namely , where
An example of signal-space diagram for 8-PSK

The transmitted M-ary QAM signal for symbol i can be expressed as
( )
( )
2 1
2
cos 2 0 , 1, 2,...,
s
i c s
s
i
E
s t f t t T i M
T M
t
t
| |
= + s s =
|
\ .
( ) 2 1
i
i M u t = 1, 2,..., i M =
s
E
2 m
3 m
4 m
5 m
6 m
7 m
8 m
Decision
boundary
2
|
message
point
s
E
s
E
d
d
M t
M t
1 m
Decision
region
1
|
s
E
( ) ( ) ( )
2 2
cos 2 sin 2 , 0 ,
n n c n c
E E
s t a f t b f t t T
T T
t t = s s
36

where E is the energy of the signal with the lowest amplitude, and
are amplitudes taking on the values, and,

where M is assumed to be a power of 4.
The parameter a can be related to the average signal energy by

An example of signal-space diagram for 16-square QAM.

4.2.1.3 IIFT and FFT
Inverse DFT and DFT are critical in the implementation of an OFDM system.

IFFT and FFT algorithms are the fast implementation for the IDFT and DFT.
In the IEEE 802.11a, the size of IFFT and FFT is N=64.signal representation
using IIFT and FFT in OFDM.

2
E a =
( )
2
, , 3 , , log 1 ,
n n
a b a a M a = .
( )
3
2 1
s
E
a
M
=
2
1
0
1
[ ] [ ]
N
j kn
N
k
IDFT x n X k e
N
t
=
=

2
1
0
[ ] [ ]
N
j kn
N
n
DFT X k x n e
t

=
=
( )
( )
1
0
Re
1
cos2 sin2 , 0,1, 2 -1.
n n
N
k k n k k n
k
s x
A f t B f t n N
N
t t
=
=
= =

37

If these components are applied to a low-pass filter at time intervals then.

4.2.1.4 Orthogonallity
Digital communication systems
In time domain In frequency domain

OFDM
Two conditions must be considered for the orthogonality between the
subcarriers.
1. Each subcarrier has exactly an integer number of cycles in the
FFT interval.
2. The number of cycles between adjacent subcarriers differs by
exactly one.

Example of four subcarriers within one OFDM symbol, Spectra of individual subcarriers

( )
1
0
1
( ) cos2 sin2 , 0 .
N
k k k k
k
s t A f t B f t t N t
N
t t
=
= s s A
( ) ( )
*
1 ,
0 ,
i j
i j
x t x t dt
i j
=

=
}
( ) ( )
*
1 ,
0 ,
i j
i j
X f X f df
i j
=

=
}
( ) ( ) ( )
1 1
2 2 2
0 0
e e
s s s s s
s s
k n n k
N N
j t t j t t j t t t T t T
T T T
n n k
t t
n n
d dt d e dt d T
t t t

+ +
= =
= =

} }
38

4.2.1.5 Guard interval and cyclic extension
OFDM symbol duration:-

Two different sources of interference can be identified in the OFDM system.
Intersymbol interference (ISI) is defined as the crosstalk between signals
within the same sub-channel of consecutive FFT frames, which are
separated in time by the signaling interval T.
Inter-carrier interference (ICI) is the crosstalk between adjacent
subchannels or frequency bands of the same FFT frame.
Delay spread
Environment Delay Spread
Home < 50 ns
Office ~ 100 ns
Manufactures 200 ~ 300 ns
Suburban < 10 us

total g
T T T = +
Guard Interval DATA
g
T T
Guard
interval
FFT integration duration
OFDM symbol duration
39

To eliminate ICI, the OFDM symbol is cyclically extended in the guard interval.
This ensures that delayed replicas of the OFDM symbol always have an integer number
of cycles within the FFT interval, as long as the delay is smaller than the guard interval.

If Tg < Tdely-spread
Tg Symbol 1 Tg Symbol 2 Tg Symbol 3 Tg Symbol 4
Tdely-spread
If Tg > Tdely-spread
Tg Symbol 1 Tg Symbol 2 Tg Symbol 3
Tg Symbol 1 Tg Symbol 2 Tg Symbol 3 Tg Symbol 4
Tg Symbol 1 Tg Symbol 2 Tg Symbol 3
Tdely-spread
Guard Interval
(Cyclic Extension)
40

4.2.1.6 ADVANTAGES and DISADVANTAGES
Advantages
Immunity to delay spread
Symbol duration >> channel impulse response
Guard interval
Resistance to frequency selective fading
Each subchannel is almost flat fading
Simple equalization
Each subchannel is almost flat fading, so it only needs a one-tap
equalizer to overcome channel effect.
Efficient bandwidth usage
The subchannel is kept orthogonality with overlap

Disadvantages
The problem of synchronization
Symbol synchronization
Timing errors
Carrier phase noise
Frequency synchronization
Sampling frequency synchronization
Carrier frequency synchronization
Need FFT units at transmitter, receiver
The complexity of computations

41

4.3 Matlab code for noise minimization in an OFDM signal
clear all
clc
close
M = 4;
no_of_data_points = 64;
block_size = 8;
cp_len = ceil(0.1*block_size);
no_of_ifft_points = block_size;
no_of_fft_points = block_size;

data_source = randsrc(1, no_of_data_points, 0:M-1);
figure(1)
stem(data_source); grid on; xlabel('Data Points'); ylabel('Amplitude')
title('Transmitted Data "O"')

% Perform QPSK modulation
qpsk_modulated_data = pskmod(data_source, M);
scatterplot(qpsk_modulated_data);title('MODULATED TRANSMITTED DATA');

num_cols=length(qpsk_modulated_data)/block_size;
data_matrix = reshape(qpsk_modulated_data, block_size, num_cols);

cp_start = block_size-cp_len;
cp_end = block_size;

% cyclic prefixing
for i=1:num_cols,
ifft_data_matrix(:,i) = ifft((data_matrix(:,i)),no_of_ifft_points);

for j=1:cp_len,
actual_cp(j,i) = ifft_data_matrix(j+cp_start,i);
end

ifft_data(:,i) = vertcat(actual_cp(:,i),ifft_data_matrix(:,i));
end

% 4. Convert to serial stream for transmission
[rows_ifft_data cols_ifft_data]=size(ifft_data);
len_ofdm_data = rows_ifft_data*cols_ifft_data;

% Actual OFDM signal to be transmitted
ofdm_signal = reshape(ifft_data, 1, len_ofdm_data);
figure(3)
plot(real(ofdm_signal)); xlabel('Time'); ylabel('Amplitude');
title('OFDM Signal');grid on;

noise = randn(1,len_ofdm_data) + sqrt(-1)*randn(1,len_ofdm_data);

42

avg=0.4;
for i=1:length(ofdm_signal)
if ofdm_signal(i) > avg
ofdm_signal(i) = ofdm_signal(i)+noise(i);
end
if ofdm_signal(i) < -avg
ofdm_signal(i) = ofdm_signal(i)+noise(i);
end
end
figure(4)
plot(real(ofdm_signal)); xlabel('Time'); ylabel('Amplitude');
title('OFDM Signal with noise');grid on;

channel = randn(1,block_size) + sqrt(-1)*randn(1,block_size);

after_channel = filter(channel, 1, ofdm_signal);

% 2. Add Noise
awgn_noise = awgn(zeros(1,length(after_channel)),0);

% 3. Add noise to signal...

recvd_signal = awgn_noise+after_channel;

% 4. Convert Data back to "parallel" form to perform FFT
recvd_signal_matrix = reshape(recvd_signal,rows_ifft_data,
cols_ifft_data);

% 5. Remove CP
recvd_signal_matrix(1:cp_len,:)=[];

% 6. Perform FFT
for i=1:cols_ifft_data,
% FFT
fft_data_matrix(:,i) =
fft(recvd_signal_matrix(:,i),no_of_fft_points);
end

% 7. Convert to serial stream
recvd_serial_data = reshape(fft_data_matrix, 1,(block_size*num_cols));
scatterplot(recvd_serial_data);title('MODULATED RECEIVED DATA');

% 8. Demodulate the data
qpsk_demodulated_data = pskdemod(recvd_serial_data,M);
scatterplot(recvd_serial_data);title('MODULATED RECEIVED DATA');
figure(5)
stem(qpsk_demodulated_data,'rx');
grid on;xlabel('Data Points');ylabel('Amplitude');
title('Received Data "X"')

43

4.4 Output and results

44

CONCLUSIONS FROM THE PROJECT
- We used the Nayantara Image(figure 1) in png format ,adding four noise
(Speckle, Gaussian ,Poisson and Salt & Pepper) in original image with standard
deviation(0.025)
De-noised all noisy images by all filters and conclude from the results that:
(a)The performance of the Wiener Filter after de-noising for all Speckle, Poisson
and Gaussian noise is better than Mean filter and Median filter.
(b)The performance of the Median filter after de-noising for all Salt & Pepper
noise is better than Mean filter and Wiener filter.
- Part of the importance of the EMD is the way it decomposes signals over which
Fourier Transforms stumble. Because of the way the EMD keeps signals in their
own domain, it can deal with signals which would otherwise be considered
"poorly-behaved." When a signal enters a new domain, the way certain select
characteristics change with the original variable are totally lost. While the signal
may be accurately retrieved, the signal cannot be effectively analyzed in the new
domain without this information.
- The EMD does not have this problem. When it decomposes a time-domain signal
into IMFs, for instance, each mode function contains information about how the
frequency of the original signal changes in time. Because of this, the EMD needs
not assume the slightest pretention towards linearity or time-invariance.
- The EMD is obviously far superior when used for the right purposes. For outputs
of linear, time-invariant systems the EMD is all but worthless, and time-
consuming to calculate besides. For non-linear, non-stationary signals like so
many signals are in the real world, however, the EMD is not only a useful method
but possibly the only computational method of analysis.
- Taking our example of the ECG, until recently readouts of test data like that
which we used were examined by eye and results were determined by estimation.
Methods like this are not standardized and not very repeatable. The EMD may
provide efficient handling of natural data.
-
- LPF filtering is a must to filter out the noise, the efficiency for noise removal is
very poor in this case. We have used a 5
th
Oder Butterworth filter to get the best
response possible.
- Least mean square (LMS) algorithm is a very efficient method for noise
minimization in communication systems. The adaptive nature of the algorithm
provides the best output among LPF, EMD and LMS.
- In emerging 4G technology in wireless communication OFDMA is the access
scheme for speed and efficiency to provide error free communication using a
OFDM channel. We have designed a code for addition of noise and its removal in
an OFDM channel.

45

FUTURE SCOPE OF WORK

- Noise minimization in video signals
- Analysis of variety of noise in image and their implementation into the
codes designed.
- Hardware implementation of the study done
- To design an algorithm based on the study in Oder to publish the research

REFERENCES

The lists of references for the project are:-
1. Freesound.org - 'avalanche.wav' by mystiscool_files
2. Reuter_Schweizer_Sound_triggering_ISSW09
3. A20_PhysRevLett_93_238001
4. acoustic-wave-technology-sensors-936
5. Chanchal De PhD Thesis
6. TA_Radio waves_EN
7. Basics of wireless communication by Theodore.s.rappaport.
8. Analog and digital communication by B.P.Lathi.
9. Analog communication by milliman halkias.
10. Digital signal processing by Oppenheim.

46

BIBLIOGRAPHY
www.wikipedia.com
www.mecanum.com/files
www.cdbaby.com/cd/ChuckWilson

www.treemangathering.com/hollowvoicesnewcd.html
www.ias.ac.in/sadhana/Pdf2007Jun/155.PDF
Houston Chronicle Archives
http://hdl.handle.net/2115/20232
www.sciencedirect.com
www2.imm.dtu.dk/~pch/Projekter/acoustic.html
www.mpl.ucsd.edu/people/pgerstoft/asa/neumann.pdf
www.iitd.ac.in/

Noise Minimization in Images and Communication

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Noise Minimization in Images and Communication

Uploaded by

Copyright:

Available Formats

MAJOR PROJECT

MAJOR PROJECT REPORT- NOISE MINIMIZATION TECHNIQUES

You might also like