Professional Documents
Culture Documents
S-DSIP Toppers 2019
S-DSIP Toppers 2019
1 --
Sl£llabllli
Exam TT-1 TT-2 AVC Term Work Oral/Pract ical End of Exam Total
Marks 20 20 20 25 25 80 150
We have tried to cover almost every important topic from above syllabus (theory only). If you feel
something is miss out and not solved in this solution, then do contact us via WhatsApp /+91993003 8388)
or Facebook (FB/Backk BenchersC ommunity) . We will provide solution for the same.
For sums, refer Last Moment Tuitions (LMT) Videos. LMT is a YouTube Channel which provide the clear
knowledg e of various subjects of Engineerin g.
musolution.in
Pagel of SS
Scanned by CamScanner
I Discrete Time Signal & Discrete Time System BE I SEM- 7
[PI Medium]
ENT VARIABLE:
Sequences are often altered and manipulated by modifying the index n as follows:
Y(n) = x(f (n))
Where f (n) is some function of n.
If, for some value of n, f (n) is not an integer, y(n) = x(f(n)) is undefined.
Determining the effect of modifying the index n may always be accomplished using a simple tabular
approach of listing, for ea ch va lue of n, the value of t (n) and the n settirig y(n) = x( f (n)).
s. However, for many index tra nsformat ions this is not necessary, and the sequence rr:iay be determined
or plotted directly
6. The most common tr;, nsform at ions include shifting reversal, arid scaling, which are defined below.
I} Shiftjng:
l. This 1s the transformation defined by f (n) = n - no,
2. Consider y(n) = x(n - no)
3. If n 0 is positive, x(n) is shifted to the right by no samples (this is referred to as a delay)
4. If no is negative, x(n) is shifted to the left by no samples (referred to as an advance).
II} Reversal;
1. Thie; is the transformation defined by f (n) = - no and simply involves "flipping" the signal )l'.(n) with
respect to the index n .
n = u. ±N. ±2N. · ··
musolution.in
• Handcrafted by BackkBenchers Community Page3of55
Scanned by CamScanner
1 I Discrete Time Signal & Discrete Time System
VI I) Reconstruction filten
1. Output signal of DAC is analog, which means it is a continuous signal.
2. But it may contain high frequency components.
3. Such high frequency components are unwanted.
4. To remove these components; reconstruction filter Is used.
ADVANTAGES;
1. It is cheaper to implement.
2. It can be stored on disk.
3. It can be easily duplicated .
DISADVANTAGES:
1. DSP hardware is more expensive than general purpose microprocessors & micro controllers.
2. Finite word length problems.
3. Bandwidth Limitations.
4 . Power consumption is high.
APPLICATIONS:
l. Filtering.
2. Image Processing.
3. Speech Recognition.
4 . Signal Ana lysis.
5. Wave Form Generation.
Table l.l shows the comparison between Digital and Analog Signal Processing.
musolution.in
• Handcrafte d by BackkBenchers Community
Scanned by CamScanner
Discrete Time Signal & Discrete Time System BEi SEM-7
I • a I •
"
-1 - 1 12 ) ◄ ~6 7 I
(al A dlacn:1<-1lme slpal .
.
-2 _, ' 12J4567
(b) A delay by no • l.
• •
fl
•••I
-I -7 -6 -,
I r 11 1c: .
-◄ -) -2 -1 I 2
(c) TI me mcflll.
• •
-2 -I • 5• 6•7•R• •
•3 4
(dl Down-.ampllng by a r.oc1or or 2.
n
• •
-7. _,
.1.l.1
345678 9
t•l Up.sampling by a racior or 2.
10 II
The most common types of amplitude transformations are addition, mu ltiplication, and scaling.
g Performing these operations is straightforward and involves only point wise operations on the signal.
~- Addition: The sum of two signals:
y(n) = x1(n)+ x2 (n) Where 00 < n < 00 is formed by the point wise addition of the signal values.
14. Multiplication: The multiplication of two signals:
Y(n) = x1(n) x2 (n) Where oo < n < 00 is formed by the point wise product of the signal values.
5. Scali ng: Amplitude sca ling of a signa l x(n) by a constant c is accomplished by mu ltiplying every signal
value by c:
y(n) = cx(n) Where oo < n < oo This operation may also be considered to be t he product of two signals.
x(n) and f (n) = c.
---- -- ---- .-- ----- ---- -- ---- -------------- ------------- -·--- ----- -------- ------.... ------- ------- --- ------ ---- -------- -- ------ -----........ ------
Ans: [P I Medium]
2. In a det ermin ist ic signal there is no uncertainty wit h respect to its value at any time.
3. E.g.: x (t) = coswt & x (n)=7.'Tlft
4. A ra ndom sig nal is one w h ich cannot be represented by any mathematical equation.
5. E.g.: Noise g enerat ed in electronic components, t ransmission channels, cables et c.
musolution.in
• Handcrafted by BackkBenchers Community Page~of55
Scanned by CamScanner
1 I Discrete Time Signal & Discrete Time System
BE ISEt,t
N 2
P= !~ (zN1+ 1) Llx(n)I
-N
2. A signal x (n) is said to be energy signal if and only if the total normalized energy is finite and
zero. I.e. O<E< 4
+ex,
Llx(n)l2
-ex,
SYSTEM:
l. A system is a set of e ements or functiona bocks that are connected together and produces a
output in response to an input signal.
2. E .g .: An audio amplifier, attenuator, TV set etc.
1. A system is said to be static or memory less if its output depends upon the present input only.
2. Static System has no storage devices.
3 Example: y (n) =4x (n) & y (n) =log x (n)
4. The system 1s said to be dynamic with memory if its output depends upon the present and past input
values.
s. Dynamic System has the storage devices.
6 Example: y (n) =x (n) + x (n-2) & y (n) =x (n) * x (n-1)
11) Linear and Non-linear systems:
1_ A system is said to be linear if superposition theorem applies tc that system.
2_ Superposition principle states that the response of the system to the weighted sum of signal is equal
to corresponding weighted sum of output of a system to individual input signal.
3 _ Example: H[a,x1(n) + a2x2(n) = a,H[x,(n)] + a2H[x2(n)]
musolution.in
w Handcrafted by BackkBenchers Community Pages of JI
Scanned by CamScanner
1 I Discrete Time Signal & Discrete Time System BEi SEM-7
4. If it dces not satisfy the superposition theorem, then it is said to be a no,n- linear system.
5. Example: H(a1x,(n) + a2x2(n) != a1H[x1(n)) + a2H[x2(n)]
ns: [P I Medium]
Convolution is important because it relates the three signals of interest: the input sjgnal, the output
signal. and the impulse respons~
Convolution provides the mathematical framework for Digital Signal Processing.
It states that for a Linear Time-Invariant system, if the input sequence x (n) and impulse response h (n)
are known, y (n) can be found out from the convolution sum.
The sum is represented as:
y (n) = x (n) • h (n). Where• denotes the convolution operation.
musolution.in
Handcrafted by BackkBenchers Community Page 6of55
Scanned by CamScanner
1 I Discrete Tirne Signal & Discrete Tirne System
eropertjei;
1. .Commutative Law.; '
a. The commutative property states that the order in which two sequences are convolved la
important.
I
x(n) y(n) h(II) J(n)
• •
I •
I
~(n) .r(n)
f
.f(it)
)'(n)
musolution.in
., Handcrafted by BackkBcmchers Cornrnunlty
Scanned by CamScanner
11 Discrete Time Signal & Discrete Time System BEISEM-7
,oRRELATION:
I. It is used to compare two signals.
2. It is very important operation in signal & image processing.
13. Correlation is the measure of degree to which two signals are similar.
14. Correlation as a mathematical operation closely resembles convolution.
5. Correlation of two separate signals is known as cross correlation.
15. While correlation of signal with itself is known as auto correlation.
a,,.,..,, - correlation:
+w
Rxy (1) = L
_..,
x (n).y (n -1)
tross Correlation:
+oo
Rxx (1) = L
-a,
x (n).x (n -1)
>ronerties:
Auto correlation is an even function .
The cross correlation is not commutive.
The result of auto correlation is maximum when signal matches with itself and there is no phase
shifting.
- - --- --- - ---------------------- ---- --- -----------------------------------------------------------------------------·- ---
Tl SYSTFM:
;.ausalitv of l Tl Svstem:
X - ) a;
/
this depends on fuhlre values of x(n)
Hence, for a system to be casual, its h(n) must be zero for n < 0.
musolution.in
Page8of55
If Handcrafted by BackkBenchers Community
Scanned by CamScanner
1 I Discrete Time Signal & Discrete Time System
A bounded signal is any signal such that there exists a value such that the absolute value of the signal·
never greater than some value. Since this va lue is arbitrary, what we mean is that at no point can thesig
tend to infinity, including the end behavior.
A- -- - - - - - - - - - - - - --
-A
---..,,.. ---,.
The continuous-time LTI system with impulse response h(n) is BIBO stable if and only if it is absolut
summable. That is:
00
I
n=-oo
lh(n)I < oo
- ---- ----- -- --- ------ --- -- ---------- ----------..------- -... -- - --- ·-------
QB. Differentiate IIR & FIR Systems
Ans:
[PI Medium)
musolution.in
• Hand.;; afted by BackkBenchers Community Page of 55
9
Scanned by CamScanner
I Discrete Time Signal & Discrete Time System BE I SEM-7
Sam p ling is d efin ed as "The process of measuring the instantaneous values of continuous-time signal
in a d iscrete fo rm ".
~ Sample is a p iece of data t aken f rom t he w h ole data which is continuous in the time domain.
~- When a source generates an analog signal and if that h as to b e digitized, having ls and Os Le., High or
Low, t he signa l has t o be discretized in t ime.
!'+. T his discretization of an alog signal is ca lled as Sampling.
;,. The following fig ure indicates a continuous-time sigr,al x(t ) and a sampled signal Xs(t).
p. When x(t) is multiplied by a periodic impulse train, the sampled signal Xs(t) is obta ined.
17. Figure 1.7· shows the example of sampli'lg
x(tl /\
r'\ A~ \
. vv.
continuousTimoslonal
\j
•• hi I I
1111 I
r· The dig itization of analog signals invo lves the rounding off of the values which are approximately equal
t o t h e analog values.
'2. Th e m ethod of sampling c hooses a few points on the analog signal and then these points are joined to
musolution.in
I
Scanned by CamScanner
11 Discrete Time Signal & Discrete Time System BE I SF.•· .
1001 , I
l . I .!
.,
OJJO
I
O!Ol • -• •• • •• -
0100 •• L".. , -· ,. I
0011 ·-·+·· - .· ◄•• .!........
I I I t I
' ... - ~ ·-
0010
0001
t-"·t ·! · l ,.j ~1- ·
·- · !"·· --!· ·t . -
0000 L...._ _ · _- ·_··-..:..
1-_-_ _--:-:-"""'."".'"-:-:--:-:--'.:-;--
0.l llOJ O.◄ O..S 0.60.70.811.91.0 II Ul..l
Figure 1.8
9. The spacing between the two adjacent representation levels is called a quantum or step-size.
10. The following figure 1.9 shows the resultant quantized signal which is the digital form for the giv
analog signal.
Figure 1.9
E - E
fl- -Q)
iz(n)f
musolution.in
• Handcrafted by eackkBenchers Community
Page·nof 5f'
Scanned by CamScanner
Discrete Time Signal & Discrete Time System
BE I SEM • 7
1 +N
p = Nlim - - " lx(n)l2
-+oo 2N + 1 ~ N
1. State each of the following statement is True OR False. Justify your answer in 4 or 5 sentences.
(a) Linear phase Filters are always IIR
(b) For a causal system h[n] tends to zero, as n tends to infinity, the system is stable.
(c) A stable filter is always causal
(d) A stable, causal FIR filter has its poles lying anywhere inside the unit circle in the z plane
(e) IIR filters have recursive realization always
s: [PI Medium]
Linear
For linea r phase filte r h [n) must be either Sym metric or Antisymmetric
For symm h[n], h [n] - h[N l m] and for antisymmetric h (n], h [n]:: - h [N-1-m].
In case of II R fi lter this cond tion Is not guarantef!d as the length o f II R filter is infinite.
H owever In case of r- lR filter, length of FIR filter is finite and so the conditio n of symmetric h[n] or
a n tisymmetric h[n) can be easily fu lfi lled.
Therefore linear phase f ilters are always FIR Filters.
Fo r causal syst em the output of the system is given by y [n) :: x [n) • h[n]
"'
y [nJ - Lx[m) "
h[n - m) - Ix(m]h(n m]
..., m-0
The condition for stable filter is ROC must include unit circle.
For causal filter, when pole< 1 system is stable.
For Non causal filer with anticausal h[n], when pole> 1 system is stable and with both sided h[n] suppose
eg. i) rnus.,I & st,,ble ii) Ant1caus.\l & st,,ble iii) Bothsided & st,,ble
(d) A stable, causal FIR filter has its poles lying anywhere ·
1. In case of causal FIR Filter, poles are always only at origin. i.e. pole= 0.
2. For causal and stable filter all the poles must lie inside the unit circle.
3. Therefore, FIR filters are always stable filter with poles only at origin.
b 0 +b 1 +·····+ bmz
H(z) = -N .
l+a 1 +····•+aNz
3. This gives difference equation in the form,
IMPULSE RESPONSE:
l. When the input to a DTS is unit impulse <'i(n) then the output is called an impulse response of
:;ystem.
2. More generally, an impulse response 1s the reaction of any dynamic system in response to so
external change.
3. It is given as:
h(n) = H[o(n)]
= c'i(n) • h(n)
o(n)
Input Signal 1
J{
J
Discrete Time System
• h(n)
Output Signal
musolution.in
., Handcrafted by BackkBenchers Community
Page1Jof
Scanned by CamScanner
11 Discrete Time Signal & Discrete Time System BE I SEM-7
:;rep RESPONSE;
11 ,
1. When the input to a DTS is unit step signal u(n) then the output is called an step response of the system.
12. It is given as:
s(n) = H[u(n)]
s(n) = u(n) • h(n)
h(n) = s(n) - s(n-1)
~ ... .,. __ ---- - ---- --- -- - -- ------- ----- ----- -- ------- ------- --- - --- -- -- ------ -- -------- -------- -- ------------ -- ------- ... ------ ----------- -- -- ---
y(O) ±
m•O
x[m] ll[ml - ( J1 ( I ) (O)(3) + ( 0 )( 2 )+ ( 1 )( I ) - [D
y[l I "f
__, xfm]ll[m-1 1 -= ( l )( . )+( 0 )( 4 )-r (0)(3) + (1)(2) = IT]
111 =0
mko
x[m] b(ru-3] (I )( . ) + (0)(.)+(0)( . )+ (1)(4) = 0
Step - 2: To find y[n) for n < 0
3
•• L
y[n] = L x[m] b[m-n) x[m] h[m-u]
---•., m=O
y[-1) = L
M=O
x[m]h[m+l] = (1)(3)+(0)( :!)-r(0)(.)-(1)(.) = 0
y[-.!) = L3 x[m]h(m-r2) = (1)( .:!)+(O)l.)-r--(0)( . )+ (!)(.) = DJ
m=O
---... - ----- --- ---------- --- ---------- -- ------ -------- ------ --- -- ----------- -- ---- -......... -------- --------- --- .. ---
,.
_.. - ..
musolution.in
.._
Page14of 55
Scanned by CamScanner
1 I Discrete Time Signal & Discrete Time System
musolution.in
• Handcrafted by BackkBenchers Community
Page1sof'
Scanned by CamScanner
I Discrete Fourier Transform BE I SEM-7
applications.
Fourier analysis is named after Jean Baptiste Joseph Fourier (1768-1830), a French mathematician and
physicist.
Fourier Transform is one of the most important transforms that is used in Digital Signal Processing &
Image Processing.
The Fourier representation of periodic discrete time signals has been extended to non-periodic signals
~OPERTIES OF OJ:T:
ax(n) aX(k)
Scaling
a 1x,(n) + a-y::.(n) a:X1(k) + a'J!'((k)
Linearity
x(n) =x(n + N) X(k) =X(k+ NJ
Periodidty
X(k)e -jl;rkm!N
Circular Tune Shift x((n - m))N
x(n)eflm1m/N X((k-m))N
Circular Frequency Shifi
x((- n))N = x(N- n) X((-k))N-X(N-k)
'Time Reversal
x 1 (n) 0 xz(n) X (k) .Xlk)
Circular Convolution 1
1
Multiplication x 1(n) . xi(n) N [X 1 (k) @Xi(k)]
N-1
L lx{n)l 2 :L k-O
IX(k)l
2
n-O
X *(n) X\N- k)
Complex Conjugate
musolution.in
Pagel& of 55
rnmmunity
Scanned by CamScanner
2 I Discrete Fourier Transform
LINEAR CONVOLUTION;
l. Multiplication of two sequences in time domain is called as Linear Convolution.
2. Linear Convolution is given by the equation y(n) = x(n) • h{n) & calculated as
co
y(n) = L
-co
x(k)h(n - k)
3. Linear Convolution of two signals returns N-1 elements where N is sum of elements in both seque
Basic operations:
1. Figure 2.1 sh ows flow diagram
h[tt]3==:> N-pointDFT
:i:[n]~ Extend to N-
polnt ==:!)> N-poinl OFT
Example:
Given:
x [n] = 6 [n] + 6 [n-1] - 6 [n - 2] + 6 [n - 3] + 6 (n - 4] + 6 (n - S]
h (n] = - o [n ] + 26 [n - 1] - 6 [n - 2]
musolution.in
• Handcrafted by BackkBenchers Community '-
Scanned by CamScanner
I Discrete Fourier Transform BE ISEM-7
CXI
""
(vii) n - 6,y(6) - I ><m)/,(6 - 111)
11!• 0
y[6] = (1) (- 1) + (1) (2) + 0 ,. J.
(I)
herefore,
y[n) = { t l , 1, 2, - 4, 3, 0, 1, - 1 I
----------------------------------------------------------------------------------------------------------------------------------
Explain circular convolution using OFT
[PI High)
N-1
y(m) = L xi (n) x2 (m-n)N
n=O
Circular convolution returns same number of elements that of two signals.
...a'..,!.j''i..:!11!E
p.a"""u.T".lLJu CIRCULAR CONVOLUTION:
) Concentric Circle Method:
Let xl(n) and x2(n) be two given sequences.
musolution.in
Handcrafted by BackkBenchers Community Page18ot•
Scanned by CamScanner
2 I Discrete Fourier Transform
N ow we take t wo concentric circles. Plot x,(n) on the outer circle in the counter clockwise direction a
i.e.
X2(n) on the in ner c ircle in the clockw ise direction.
,.
We multiply t he corresponding samples and add. This gives us y(O)
:. y (O) =4x • 2x2 +1 x3+ l x 4
= 15
We now rotate the inner circle in the anti-clock wise d irection by o ne st e p . This g ives us,
:. y(l) =l x l + 4 X2 + 2 X 3 +l X 4
= 19
We continue rotati ng t h e inner circle in st eps of one.
:. y(2) = l x l + l x2+4x3+2x4
= 23
:. y(3) =2x l + l x2+ l x3+4x 4
= 23
If we now rotate the inner c:ircle, we com e b ack t o t h e starting point. He n ce w e stop here.
:. y(n) = {15, 19, 23, 23}
1. M at rix method represents t he t w o given seque nce xl(n) and x2(n) in matrix form .
2. O n e of t he given seq uences w ill be repeated v ia circular shift of one sample at a time to form a N )(
m atrix.
3. O ther sequence will be represented as column matrix.
4. Multi plicat ion of two m at rices gives the result of circular convolution
5., f;=:xa':;"ple:
musolution.in
• Handcrafted by BackkBenchers Community
Scanned by CamScanner
musolution.in
Scanned by CamScanner
3 I Fast Fourier Transform
.EEn
l. FFT stands for Fast Fourier Transform.
2. A fast Fourier transform (FFT) is an algorithm that computes the discrete Fourier transform (DF'tl
sequence, or its inverse (IOFT) .
3. Fourier analysis converts a signal from its original domain (often time or space) to a representatl
the frequency domain and v ice versa.
4. The Fast Fourier Transform (FFT) is simply a fast (computationally efficient) way to calculate
Discrete Fourier Transform (OFT).
FFT WORKING:
1. By making use of periodicities in the sines that are multiplied to do the transforms, the FFT g~
reduces the amount of calculation required.
2. Functionally, the FFT decomposes the set of data to be transformed into a series of smaller data
be transformed.
6. For exa m ple, an FFT of size 32 1~ b roker into 2 FFTs of size 16, which are broken into 4 FFfs ofs
which are b roken into 8 FFTs of size 4, which are broken into 16 FFTs of size 2.
7. Calculati ng a OFT of size 2 is trivial.
FFT EXAMPLE:
l. The main advantage of having FFT is that through it, we can design the FIR filters.
2. Mathematically, the FFT can be written as follows;
N-1
x[K] = L
n=O
x(n) w;k
3. We have considered eight points named from xo to x,.
4 . We will choose the even terms in one group and the odd terms in the other.
S. Diagrammatic view of the above said has been shown in figure 3.1.
XO Xl X2
Figure 3.1
musolution.in
• Handcrafted by BackkBenchers Community
Scanned by CamScanner
I Fast Fourier Transform BE I SEM-7
Here, points xO, x2, x4 and x6 have been grouped into one category and similarly, points xl, x3, xS and
'
x7 has been put into another category.
Now, we can further make them in a group of two and can proceed with the computation.
Now, let us see how these breaking into further two is helping in computation .
.!!._1 !!._1
2 2
x[k) = L x[2r) wt1c + L x[2r + 11wi2r+1)lc
r=O r-0
N N
~2- 1 [2 ]Tvrk ~2-1 [2 wkN
= L.,,--o X r ., N/2 + L.,,-.-o X r + l)Wrlc
N/2 X
= G[k] + H[k] x Wt
Initially, we took an eight-point sequence, but later we broke that one into two parts G[k] and H[k].
G[k] stands for the even part whereas H[k] stands for the odd part.
If we want to realize it through a diagram, then it can be shown as below in figure 3.2
G[OJ
G{1]
Figure 3.2
w~~ 1
w;- -wi
w: - wi8
w: - -w:
musolution.in
r.
Page22of 55
Handcrafted by BackkBenchers Community
Scanned by CamScanner
3 I Fast Fourier Transform
15. The disadvantage of this system is that I< cannot be broken beyond 4 point.
16. Now Let us break down the above into further.
17. We will get the structures something like this
4. Though it is not the efficient algorithm, it lays foundation for time-efficient OFT calculations.
5. Let t h e N point data sequence x (n) be splitted into two N/2 point data sequen c es f (n) and fi{n) s
1
that f 1(n) contain even number of sequence and f2(n) contain odd number of sequence.
6. Therefore, we can write that
/, (n) • ..,(2n)
n• O. I, 2, ....... ,.t;_ -1)
/2 (11) = x(2n + 1) 11 ., o. I • 2.· ..... ... N - I
2
7. Above equations shows that the time domain sequence is splitted into two sequences.
a. The above splitting operation is called decimation.
• X(k) = L
n=o
x(n)W:1'
Where k = O, 1, 2, 3 ...... N _ 1
Splitting the sequence x(n) into even and odd number of sequence, we get
X(k) =
n """"
L x(n)W,.r1 + I
n odd
x(n) w,.ri
N zN -1
X(k) = , 2 - 1 X (2m) w,;mk + ' X (2m + 1) w.<2m+t)k
Lm~ L N
In above equation, let put f1(m) =x(2m) and f 2(m) =x(2m + l).
zN -1
X(k) = I::01 N
/1(m) (WN"k)2 + L
m=O
f2(m) (WN"k) 2 w.,:
Since WJ = W!! 2
we have
X(k) = I:: N
0
1
/ 1 (m) Wfk + wt +
We know that WN
k+~)
2
= - wt
X(k + (N/2)) = F1(k) - wt F2(k) Where k = 0, l, 2, 3, ....., (N/2) - l
From above equation, we can say that X{k) is N-point OFT.
Therefore, we can determine F1(k) and F2{k) taking k= Oto (N/2) -1 because they are N/2 point DFTs.
- - - --- --- ----- - - - & - - - - -- - - - - - .. - -- - - - - - - -- - ---- - --- --- - - - - - --- ---- - - - - ... - - - -- -- - -- - - - --- - - - -- -- - - - --- -- - - - ---- -- - - --- - .. - ... - - - - - - - - - - - - - - - -
Figure 3.3
X(k) can be obtained From F1(k) and F2(k) where F,(k) and F2(k) are two 4-point DFTs.
The symbolic diagram operation has been shown in figure 3.4.
The sequences of F1(m) and F2{m) are given below:
musolution.in
P■ge24ofll
Handcrafted by BackkBenchers Community
Scanned by CamScanner
3 I Fast Fourier Transform BEl4:.a
/,(Ill)•
/2(111) •
,(2n)•{.r(0), r(2),.t(4 ),.rC6>}
t(ui +I)• {xCl),.r(3), .r(5). .1(7)}
}-!!. 2
point ~umm
I,,1m1. a(2n) I
fi(O) •ll(O)
N 'TMNQMS
}-point
f1(1) • .11(2) DFT -oblllned
1.•• bylU!lng
f1(2) •11(4) 4 - point k•0,1,2,3
OFT In Eq. 5 5.7
r,(3) • 11(11)
'2(1) • x(3)
'2(2) • x(5)
'2(3). 11(7)
Figure 3.4
l. Let us split fi(n) into the following even and odd numbered samples:
11:0. I.••-~-•, 11
n==O 1,.......... -1 j
2 l ii(n) and "li (n) are the even and odd n u mbered sequences of fi(n) containing (N/4) sa
because /i(n) contains (N/2) samples.
3. Now let us split fi(n) into the following even and odd numbered samples:
5. Earlier we have obtained X{I<) and X(k + (N/2)) from Fi(k) and Fi(k) with fi(n) and fi(n) as decl
sequences and the length of OFT was (N/2)
6 . As before, we can obtain Fi(k) and Fi(k + (N/4)) from "11(n) and 1,1
2(n). Therefore, we can write that
f",(il : l '11 fk1 + \V.~V12 (4) k a O. I, ..... N - I
T 4
7. In above equations, the (N/2) point OFT are obtained from (N/4) point DFT.
8 . V.1(1<) and \li2(1<) are the (N/4) point DFT of "11(n) and 1,12(n) respectively.
9 . Similarly, we can write the following for Fi(k) also:
,. .
" Handcraft-'!d by BackkBenchers Community
musolution.in
Page ZS
Scanned by CamScanner
... , I Fast Fourier Transform
e.· In above equations, the (N/2) point OFT are obtained from (N/4) point OFT.
,~ Vi1(.l<1 and Vi2(1<} are the (N/4) point OFT of V.zi,(n) and ½i (n) respectively.
2. Figure 3.5 s hows /,,{I<} and F2(1<} are two 4 points OFT which are shown by two separate blocks.
00~00 00~00
Figure 3.5: A Symbolic Representation of 4 point OFT using direct computation
For N= 8, Vi1fl<1 and 1,fa(I<} are (N/4) = 2 point OFTs.
' T h e symbolic diagram of o p eration has been shown in Fig ure 3.6
From Figure 3.6(a) the sequences nu(n) and 1t12(n) for 4 point fi(n) are given by,
T h ernfore, the two (N/4) point DFT can be computed separately and combining them w e will
musolution.in
Page26of 55
Commu"
Scanned by CamScanner
I Fast Fourier Transform
Figure 3.6: Point OFT of F1 (X) and F2 (k) obtained by combining two 2 point DFTs.
l. Let 'JS further split (N/4) point sequences in their even and odd parts.
2. Therefore, we will get the next stage of decimation and the sequences which will be
length (N/8) point. (N/8) = l for N= 8.
3. Figure 3.7 shows the two-point DFTs.
v11 (n), n - 0, 1
I
Yu(k) == L," 1(n)n2
hQ
bo . k = o. 1
5. For k= O,
I I
Vn(O) lv11(n)lf1 = I,1· 11(11)= vu(0)+1•11 (1)
""° 0
6. Because W2° = 1
For k= 1,
I
Vu(I) - 2,I' .1(n)nr,. l'n(O)Jl'lo ~ Vu(l)W}
"',j
w<_!1 Wl-
We know that and 2 - -l . Now we can write as
· w 0 =1 . . .
Since 2 , let us rewrite above equations 1n the following forms:
musolution.in
• Handcrafted by BackkBenchers Community
Scanned by CamScanner
I Fast Fourier Transform BEi SEM-7
V,i(I)
r:=%{] Compulallan of
~OFT
v,,111
Figure 3.8
Figure 3.8 shows signal flow graph for third stage. It cannot be decimated further because it is a two-
point OFT.
The complete signal flow graph which is obtained by interconnection of these individual stage wise
signal flow graph is shown in figure 3.9
2-pomtOFT Combine 2-poinl OFT• This blodt_,.._ two
4 polnl OFTt, F~ and F,(--,
l ·••·····••···-··--·--·····1 F1(0)
X(O)
! WO2 •WO
1
: :
: : v,2(1) : F,(3)
-'{6) + --0.::....-- -~H-..!!::.::.__ . ! . , - -~------b....:,:..:....:.!:::!....-+---...:>--~t..._4~~~--o_;_ JC13)
:...................... .. .......... _ .... _., ___
-1 :• '• -1 •
- - - - - - - - .................................. .... ..1
!:
...
X(5)
:
.................................... .. -- -1....'
•• • ~ Vu(O) !Fi2)
.l(JJ -~ ---c---~->-!----~--Q~---::~----Q-,---~----~------.:------<i,.....,...- Xl&l
.
:'
: V22(I)
...:---o::...,__ __ ....::0--+-.:;;;.;...;....._,__-(1-_ ;_ _-:-_-i)_ _ _ _ _ _ _-t,,_Jtl7)
F,(3)
I;-~~==;;~~•J L--~·:::;=~-··'
>'(7) _ _ _ __
I
~
········;::::;;::=··-•.
Shuffled e,ray lrl
bit revensed Older
gure 3.9: Signal flow graph and stages of Computation of Radix-2 DIT OFT algorithm for N = 8.
musolution.in
Handcrafted by Backk8enchers Community Page28of55
Scanned by CamScanner
4 I Digital Image Fundamentals
1. Digital image processing (DIP} deals with manipulation of digital images through a digital corn
2. It is a subfield of signals and systems.
3. It focus particularly on images.
8. The input of that system is a digital image and the system process that image using
algorithms, and gives an image as an output.
9. The most common example is Adobe Photoshop.
10. Figure 4.1 shows the example of DIP.
I
Captured by
Send to
Gives Output as
3. Digital Image Processing do image enhancement to recollect the data through images.
4. It is used widely everywhere in many fields.
5. It reduces the complexity of digital image processing.
6. It is used to support a better experience of life.
ADVANTAGES:
DISADVANTACES:
1. It is very much time-consuming.
musolution.in
• Handcrafted by BackkBenchers Community
Scanned by CamScanner
4 1Digital Image Fundamentals BE I SEM-7
n. sampling and Quantization are two important processes use to convert continuous analog signal into
digital image.
4. Sampling corresponds to a discretization of the space or spatial co-ordinates. That is of the domain of
the function, into F: [l ....N] • [l .... M] ➔ Rm
l
... f(l,M) l
f = f (2,1) : [(2,2)
. f(2, M)
.
(N, 1) f (N, 2) f(N, M)
6. The smallest element resulting discretization of the space is called as picture element i.e. Pixel
7. For 3-0 image this is cd lled as voxel (Volumetric pixel).
B. Quantization corresponds to a discretization of the intensity va lue or grey level values. That is of the
domain of the function, into F· [l .... N] • [l .... M] ➔ (0 ... L]
9. Quantization corresponds to transformation Q (f)
4 levels 8 levels
f T
3 ),~
V/'.,
0 ill N N 1N llt IIC 1:1<
musolution.in
., ~-~ by BackkBenchers Community Page30ofSS
Scanned by CamScanner
41 Dlgltal Image Fundamentals
9. For example, for 2 bit pixel values, total no. of brightness levels. 2 3 = 8.
10. Similarly, for 8 bit pixel values, total no. of brightne!>s levels, 2 8 - 256
----------------------------------------------------------------------------------------------------------------------------------
Q3. Explain Image File Format
Ans: [P I Medi
JPEG:
1. JPEG stands for Joint Photographic Experts Group.
2. It is used to display photographs and other continuous tone images in HTML documents over w
3. It supports CMYK, RGB & Greyscale color modes.
4. It does not support alpha channel.
5. It's a standard image format for containing lossy and compressed image data.
6. Despite the huge reduction in file size JPEG images maintain reasonable im<1ge quality.
7. This unique compression feature allows JPEG files to be used widely on the Internet, Computers,
Mobile Devices.
8 . The sharing of JPEG images is quick and efficient.
9. Also, a large number of JPEG image files can be stored in minimum storage space.
10. JPEG files r:an also contain high-quality image data with a lossless compression.
D..M.ei
1. The BMP extension represents Bitmap Image file.
musolution.in
• Handcrafted by BackkBenchers Community
Scanned by CamScanner
pigital Image Fundamentals
41
__.-------------------------------~B~E~l~S~E~M~-7~
2. It is a standard windows image form at on DOS and windows- cornpatible computers.
• It supports RGB, Indexed Color, Greyscale and B't
1 map co 1or modes.
3
It does not support alpha channel.
4.
5. sMP file contains raster graphics d at a which
• are .independent of display devices.
. That means a BMP image file can be viewed w·ith ou t a graph'1cs a d apter.
6
7, BMP images are generally uncompressed or compressed with a lossless compression method.
s. The files can st0 re two-dimensional digital images with both monochrome and color.
Df.Ei.
1. TIFF stands for Tagged Image File Format.
2. It is used to exchange files between applications and computer platforms.
~- It is a standard file format that is largely used in the publishing and printing industry.
4. The extensible feature of this format allows storage of multiple bitmap images having different pixel
intermediate files.
:i. A TIFF file uses the file extension ".tif" or ".tiff''.
t7. TIFF is also known for its flexibi lity and has more features and capabilities than other image file formats.
S. Most image editing software or applications are capable of working with TIFF files.
CONNECTIVITY:
1. In binary valued digital imaging, a pixel can either have a val ue of 1 (when it's part of the pattern), or o
(when it's part of the background) i.e. there is no grayscale level.
2. We will assume that pixels with value 1 are black while zero valued pixels are whit:?.
3. In order to identify objects in a digital pattern, we need to locate groups of black pixels that are
lY.e.E.S.
I)
1 !t - Connectjyit'l: . · I p if Q and P share an edge.
· A pixel Q is a 4-neighbor of a given pixe' '
1
2 ' ' . • els P2, P4, P6 and PB) are shown in Figure 4.2.
· The 4-neighbors of pixel P (name 1Y pix
-----------------=-=--=-==-
musolution.in
pageJ2ofSS
" Ha- · ~- · ._. aackkBenchers community
Scanned by CamScanner
4 I Digital Image Fundamentals
3. A set of black pixels, P, is a 4-connected component if for every pair of pixels p, a~d PJ in P, there
a sequence of pixels pi, ... , pj such that:
a. All pixels in the sequence are in the set P ie. are black, and
b. Every 2 pixels that are adjacent in the sequence are 4-neighbors.
Figure 4 .2
Examples of 4-connected patterns:
The following figure 4.3 are examples of patterns that are 4-connected.
Figure 4.3
II} 8 - Connectivity:
1. A pixel, Q, is an 8-neighbor (or si mply a n eighbor) of a given pixel, P, if Q and P either share an P.dg
a vertex.
2. The 8-neighbors of a give n p ixe l P make up the Moore neighborhood_of that pixel.
3. A set of black pixels, P, is an 8-connected component (or simply a connected component) if for
pair of pixels p , and PJ in P, there exists a sequence of pixels p 1, ... , PJ such that:
a. All pixels in the sequence are in the set P i.e. are black, and
b. Every 2 pixels that are adjacent in the sequence are 8-neighbors
4 . All 4-connected patterns are 8-connected i.e. 4-connected patterns are a subset of the set
connected patterns.
s on the other hand, an 8-connected pattern may not be 4-connected.
Figure 4.4
musolution.in
., Handcrafted by BackkBenchers Community
Scanned by CamScanner
Fund.......
Figure 4.5
An~ ~I~
IMACE ENHANCEMENT;
l. Image enhancement is the process of adjusting digital images so that the results are more suitable
display of image or further image ana lysis.
2. For example, you can remove noise, sharpen, or brighten a n image, making it easier to identify
features.
3. Image enhancement includes gray level and contrast manipulation, noise reduction, edge sharpeni
filtering, interpolation etc.
4 . The enhancement process does not increase the inherent information content in the data.
5. But it does increase the dynamic range of the chosen features so that they can be detected easi ly,
3. In this output image pixel value is obtained directly from proc essing the in p ut im age pixel value.
4. For every input pixel value, transformati o n funct ion g ives correspon ding o u t put image pixel value an
no memory location is required t o st n re inte rmediat e re<;u lts.
5. It includes.
a. Digital Ne gative
b . Contrast stretching
c. Thresholding
d . Grey Level Slicing
e. Bit Plane Slicing
f. Dynamic Range Compression
g. Power Law Transformation
b . Smoothing Non-Linear Filters: Non- Linear Filters are a lso ca lled as Ordered Statistic Filters
Example is Median, Max and M in Filters.
C First Order Derivatjye Filters: Example is Robert, Prewit, Sobel and Fri-Chen Fi lter
d . Sharpening Second Order Derivative Filters: Example is Laplacian Filter, High Pass Filter and Hi9~
Boost Filter.
musolution.in
¥ Handcrafted by BackkBenchers Community
Page3sof5S
Scanned by CamScanner
t,ancement in Spatial Domain BE I sEM - 7
e ~,,
1t11'g
I ,n rocess·n ·
ra •
· he modification of input imag h.1st0 • eon
1ves t e gram so as to improve the visual quality of ,mag
11
,ri"0
device.
diSplaY
1udes:
/tlfl~
,. • It is a process that attempts to spread out the gray levels in an image so that they are evenly
distributed over entire range.
• The histogram of the output image is almost uniform over the entire range of gray levels.
• It provides only one type of output.
b, ~ a m Specification:
• It is a process that attempts to spread out the gray levels in an image as per the specified image
histogram.
, tv1odifies histogram of the input image closely matches with the histogram of the specified
1rnage.
c. ttjstogram Stretching:
, It is the process that attempts to spread out the gray levels in an image linearly as per the
required range of output image histogram.
__,.____ -- ------ ------ ----- -- --- --- ------... -... ----- ... --- ---- -------- ----- ---- ------ --- - --- ----------------------- ---- -- --- ------------
[PI High]
~RANSFORMAT!ON_l
Gray level transformation is a significa nt part of image enhancement techniques which deal with
. s 1· overall Graph
musolution.in
Figure · ·
Page3&6f 55
Scanned by CamScanner
BE I SEM -'I I
s I Image Enhancement in Spatial Domain
I) Unear Jransformationi , .
1. The linear transformation includes identity transformation and negative transformation.
_ In identity transformation, each value of the image is directly mapped to each other values of tht
2
output image.
3. Figure 5.2 shows the example of Identity transformation.
255
127
0
127 255
image.
6. Figure 5.3 shows the example of negative transformation .
. 'ega11ve
Transformauon
musolution.in
• Handcrafted by BackkBenchers Community· Page39-
Scanned by CamScanner
inh■ncement in Spatial Ooma1n
.
BEi SEM-7
InvLog Log
transfo rm ation.
Formula: s = e r "y
i Here, y is gamma, by which this transformation is known as gamma transfo rm at ion .
------------------------------- - - - -
-------------------------
~~-------------------------
Explain zero Memory point operations [PI High]
"'s:
~ MEMORY POINT OPERATIONS:
. oint processing.
1t 1s also k nown as P . used. i.e. Tis 1 x 1 operator.
l . ration s, single p1xe 1s are
In Zero M emory point ope th operator T and the present f(x, y)
l· It means that the new v a ue f(X' y) depends on . e function gives corresponding output image pixel
I
~ . . el value, Transformation
!=or every input image pr x . termediate results.
. . required to store in .
Value, no m em ory locat1on 15 output image pixel value.
S . . el value and 5 denotes
musolution.in
Let r d enot es input image p ix
•
PageHofM
,.
Scanned by CamScanner
5
I Image Enhancement In Spatial Domain , function.
6 ·on Tran Sformat1on
· Then S = T(r): where Tis any Zero memory Point opera t I
3. . .
D1g1tal Negative reverses the gray scale O f ·input image. . versa
4. That 1s . be now white and vice
. black in the original image will .
S. It is defined as S = T(r)
-
C reyl.evel
.
/- - - - - ; u !s
,•·
..•········....
..•·
. .•·· · !:
/~=T(r) I
.~
j
1::,••_·
10. Here Lis the number of grey levels. (256 in this case)
II) Contrast Stretching:
l. Many times we obtain low contrast images due to poor illuminates or due to wrong setting ofth
aperture.
2. The idea behind contrast stretching is to increase the contrast of the images
portions darker and the bright portions brighter.
s. As is evident from the figure, we make the dark grey levels darker by assigning a slope greater
one
6. One can assign different slopes depending on the input image and th • •
w •••··•····•••·•·•······ •..•·· 1
•"./~
Modified
Crey Level
s S•Tto ·r··· I .......,,............~.
. rn: :
v ····""···· Ir\ s,1: :
../ :a : b
: : :
~
Figure s.s
musolution.in
• Handcrafted by eackkBe:ichers Community
Scanned by CamScanner
ement in Spatial Domain
~
~rerne contrast stretching yields Thresholding .
..,snown in figure 5.5, in contrast stretching, first and last slope are made zero and the center slope Is
;ncreased.
wewauld get threshold transformation and threshold function Le. r, • r,,s, • o ands,• L • 1
figure 5.6 shows the example ofThresholding.
rhe formula for achieving Thresholding is as follows:
if r <= a
s :0
:: L-1 if r > a
k
Where Lis the number of grey levels.
n in an X-Ray or a CT Image
In such c,ccumstances, we use a transformation known as grey level slicing
musolution.in
• background. Page40of 55
~ unit
Scanned by CamScanner
Image Enhancement in Spatial Domain BE I SEt.t •~
1 - - - - - - - - - - - - - - - - - - - - - -------...~ 6
This technique of retaining the background Is called as grey level slicing with background. • 1,..
' be implemented using the formulation
,. This can
s = L-1 if a<= r <= b
=r otherwise
,
I
l·l ••· ....... ,r--...- L . 1 ......... _ _,.._......... /1
...
.•· · I
S •T(r)
.// i
.... ~
S • T(rl ,••••·•'
Identity
Modlned ,•'
,•• Modlfted ! Tran.formation
..••'.•·
Crey level Orey Level
I I i
,•
,• ,,••••••' a b \
a
Ii
b :
0 O Original Orey Level L• 1
Original Orey level l •1
r r
Figure 5.7
V) Bit plane slicjng: j
1. In this technique, we find out the contribution made by each bit to the final image.
2. Image 1s defined as a 256 x 256 x 8 image. I
3. In this, 256 x 256 is the number of pixels present in the image and 8 is the number of bits requiredtci I
represent each pixel.
musolution.in
• Handcrafted by 'eackkBenchers Community Page41 ot55 '
....
Scanned by CamScanner
ncement in Spatial Domain
Figures.a
(actor.
• BY changing the value of gamma, we obtain a family of transformation curves.
; Non linearities encountered during image capturing, printing and displaying can be corrected using
0
gamma correction.
Hence gamma correction is important if the image needs to be displayed on the computer.
• The power law transformation can also be used to improve the dynamic range of an i rnage.
tSTOGRAM:
Histogram is a graph ica I represent• tion of the intensity distribution of an image.
. In simple terms, it represents the number of pixels for each intensity value conside,ed.
tSTQGRAM
HistogramEQUALIZATION:
Equalization is a computer image processing technique used to improve contrast in
images.
Histogram equalization is used to enhance contrast.
It is not necessary that contrast will always be increase in this.
~. There may be some cases were histogram equalization can be worse.
Figure s.9
-----------------------------
musolution.in
Page42of55
rnrr,-.1nlty
Scanned by CamScanner
atial oornain s.10.
5 I lmage Enhancement In Sp •n figure
beloW'
shown
. . e has been
7. The histogram ofth1s ,mag
f Example image.
·stograrn o
Figure s.10: H 1
rzation to it. · I · th· ·
8. Now we will perform histogram equa I tion) of all the p,xe 5 ,n Is image.
obability mass func .
9. First we have to calculate the PMF (pr . tion) according to gray levels.
·ve distributive func
10. Next step is to calculate CDF 1cumulat, d ks like this.
h t the CDF calculate Ioo
11. Let's for instance consider this, t a
~,•. ':}:.i, CDF
cray:tevet Value
0.11
0
1 0.22
2 0.55
3 0.66
4 0.77
5 0.88
6 0.99
7 1
12. Then in this step you Nill multiply the CDF value with (Gray levels (minus) 1) .
13. Considering we have a 3 bpp image.
14. Then number of levels we have are 8. And 1 subtracts 8 is 7. So we multiply CDF by 7.
15. Here what we got after multiplying.
Cray Level Value ·:- CDf.
0 0.11
CDF • (Levels-1) :;-
1 0
0.22
2 1
0.55
3 3
0.66
4 4
0.77
5
0.88 s
6 6
0.99
7 6
l
7
1
r---..!:2:___ -~-+--- --;4~---J
r----z..3__ -~-+--- -~6=-----J
r---...:4;___ - - - - 8
5
6
7
musolution.in
....
Scanned by CamScanner
f9' ~,,t,ancement in Spatial Domain BE I SE"' • 7
11istograrn.
-------- ------------ --------------- ------- --------------- --- -- --------- -------------------- -- ----- ----- -- - --- ---- -- -- -- -------
i Explain Histogram Stretching
as: [P I Medium]
ISTOO AM STRECHINO:
---------:-----------'~
musolution.in
Page 44of 55
'"'-ndcratted by eack-kBenchers Community
Scanned by CamScanner
S I Image Enhancement In Spatial oomain
st:: 225. .
. . ge contra he contrast 1s
7. Now we calculate contrast from this ,ma · . to increase t
. of the ,mage
8. The formula for stretching the histogram
l(,r.y)-l'ntin • 2bpP
g(x,y) ::: ~
. intensity multiply by levels of
1 9
. . and maximum p1xe ~.
9. The formula requires finding the minimum
ray are 256.
10. In our case the image is 8bpp, so levels O f 9
225
11. The minimum value is o and the maximum value is ·
f(x,y)-0
g(x,y) = 255 - 0 * 255
f(x,y)
g( x,y) =- • 255
255
g(x,y) = f(x,y)
7 That means the output image is equal to the proc esse d image.
•
8 . That means there is no effect of histogram stretchin
g has been done at th·1s .image.
,
--------:-----=-=--~-
musolution.in
.. Handcrafted by BackkBenchers Community
Scanned by CamScanner
,.. enhancement in Spatial Domal
sl 1rflh n
e,cplain Neighborhood Pixels Processing
[PI Medium]
,.,,s:
~ o o o PIXELS PROCESSING;
,. t is also spatial domain technique in image en h ancement.
I
z. ere' we consider one pixel at a time & m o d.fy
1 1 ·t accord"1ng Iy.
H
its neighboring pixels are also taken in consideration.
3-
4. so, we change pixel value based on a neighbors.
s. ,Along with 3x3 neighborhood, SxS & 7x7 can also be used.
, . A 1ot oft hi ngs can be achieved by neighborhood processing which are not possible by point processing.
W2 W3
W1
(x -1) g(x-1,y) g(x-1, y + 1)
g(x-1,y-1)
ws W6
W4
X
g(x, y) g(x,y+l)
g(x, y-1)
wa W9
W7
(X + 1) g(X + 1, y) g(x + 1, y + 1)
g(X + 1, y-1)
filtering
9. we can perform low pass filtering (LF), high pass filtering (HF) & band pass filtering using
neighbourhood operations.
· n o
10. In 1D signals if2 signals represent voltage then, how fast the signal changes is indicatio ff requency.
ll same concept is applied to images where we have gray levels instead.
12. If gray scale change slowly over a region then LF area. E.g. Background.
13. If gray scale change abruptly over a region then HF area. E.g . Edges, Boundaries.
musolution.in
Page 46 of 55.
- Hanr.tcrafted by eackkBenchers community
Scanned by CamScanner
5
I Image Enhancement in Spatial Domain
sPAJIAL EILTERINC;
1. The use of spatial masks for image processing is called spatial filtering and the masks are ca11
filters. · ~-...
e<tssi-..
2. Spatial filtering involves passing a weighted mask or kernel over the image and replacing the
· h th es um of original
image pixel value corresponding to the center of the kerne I wit · · :"111
Pixel vah,1 Oflat.....
reg ion corresponding to the kernel multiplied by the kerne I weig
. ht· es •I\"'...-
3. For example consider digital sub-image F and 3 x 3 filter mask was given below:
6 . The mask is then moved to the next pixel location in the image and the process is repeated.
7. This continues until all pixel locations are covered.
TYPES OF FILTERS:
1. Smoothing filters are used for blurnr ,g and for noise reduction .
2. Blurring is used as preprocessing such as removal of sma ll details from image
3. Noise reduction is blurring with linear or non-linear filter.
4. Smoothing filter can be linear and non-linear.
a. Smoothing Linear Filter:
■ The output of a smoothing linear spatial filter is simply the average of the pixels contained in the
neighborhood of the filter mask.
These filters sometimes are called averaging filters or low pass filters.
• It is used to remove noise present in the image
•
Noise is normally a high frequency signal and low pass filtering eliminates the noise.
• The major usage of average filter is reduction of irrelevant detail in an image .
• Figure 5.16 shows 3x3 smoothing filter masks
R = (1/9 ) 1:7=1 z,
■ At the end of filtering the entir;. i~age is divided by _
9
musolution.in
• Hand erafted by BackkBenchers Community
Scanned by CamScanner
-----~~~~==~~--~=--------------~~~~
SI
,rnage Enhancement in Spatial O
•
• so m x n is equal to 1/mn.
Thus the coefficients pixels ar
.
oma,n
2. There are many applications, such as electronic priming, medical images, military systems th at use
sharpening technique.
Sharpening filters that are based on two derivatives. i.e. First Order Derivatives and Second Order
Derivatives.
Robert Sobel
Ox ·1 ·2 •1 Oy -1 0 1
0 0 ·2 0
1 2
0
1 ., 0
2
1
Prewlt Fri-Chen
., ., Oy ., 0 1 Ox ., .,rz ., Oy ., 0
Ox ·1
., 0 1 0 .,fl
1
../i
0
1
0
1
0
1
., 0 1 1 .rz
0 0
1 .,
0
0 1
-
b. second derivative:
• Must be zero in constant areas.
-1 -1 -1 1/9 ·1 ·1 ·1 1/9 ·1 -1 ·1
0 -1 0
-1 8 -1 ·1 e ·1 -1 9k-1 ·1
-1 4 -1
-1 -1 -1 ·1 ·1 ·1 ·1 -1 ·1
0 -1 0
musolution.in
hers community Pas,e48of55
" H•ndc:rafted by eackkBenc
Scanned by CamScanner
6 I Image Segmentation
·1
. _, ·l
-1 8 -1
-1 -1 -1
s. The idea is that gray level of an isolated image will be quite different from gray level ofls
neighbors.
-1 -1 -1 2 -1 2 -1 2 •l •1
-1 -1
-1 2 -1 -1 2 -1 -1 2 -1
2 2 2
- I--
-1 -1 -1 2 -1
+45°
-
-1 -1 2
Vertical
·1 -1 ·1 2
Horizontal
musolution.in
fl Handcrafted by eackkBenchers Community
Scanned by CamScanner
,nage Segmentation
61 1
----=:=====----------------------~B~E~l~S~EM~•:!_7
111) .Edge Detectlom
1. It is the most common
approach ford t I . .
2. Edge Detection charact e ect on of discontinuities 1n grey level.
I
Ed . er zes th e object boundaries
3. ge point can be thought f ·
O
It · th b as pixel location of abrupt gray levels.
4. 1s e oundary between t . .
Th wo regions with relatively distinct gray levels properties.
5. ere are two types of ed
ges. They are step and ramp edge.
6. Step edges are detected .
using first order derivative filters like Robert, sobel, frichen and prewit.
7. Ramp edges are detect d .
e using second order derivative line Laplacian filter.
IIJ!d on similarit.Y.:;,
1 In this approach the principal . . . • · ·
· approach 1s threshold, region growing and region splitting and merging.
2. In this approach segmental' · d
' ion 1s one by finding the regions directly.
Let R represents the entire imag . h •• R· , , b
e region segmentation as a process t at part1t1ons into n su
reg ions Rl , R2, R3 ... Rn such that,
a. Rl U R2 U .... U Rn = R
e. If p• is similar to P
f. Then p is added to R & Neighbor pixel of p• is added t o F
musolution.in
hers community PageSOafSS
Handcrafted by eackkBenc
Scanned by CamScanner
BEi
6 I Image Segmentation
bd'vide
1 that quadrant into sub-qu
5. If Predicate (region) = false for any quadrant th en su, i.===
and so on.
Algorithm: ·
. R.1when predicate 1 -- f aI se.
(R")
a. Split into four disjointed quadrants any region . k) _ t
· h redicate (RJ U R - rue.
b. Merge any adjacent regions RJ & Rk for whic P
. 1·tt ' g is possible.
c. Stop when NO further merging or sp 1 1n
-----------------------------------------------------------------------------------------------------------------------........
Q2. Explain image edge detection using Previtt Mask
[PIM. . . .
Ans:
Vertical Direction:
1. Above mask will find the edges in vertica l direction and it is because the zeros column in the vertfr:11
direction.
2. When you wi ll convolve this mask on an image, it wi ll give you the vert·1ca 1e d ges •1n an •image.
3. When we apply this mask on the image it prominent vertical edges.
4. It simply works like as first order derivate and calculates the d"ifference of pixel
. . . . ·1n a ~
1ntens1t1es -~
region.
5. As the center column is of zero so it does not include the or' • 1
. . igina values of an image but rat
hef It
calculates the difference of right and left pixel values around that edge.
6. This increase the edge intensity and it becomes enhanced com .
parat1vely to the original image.
Horizontal Direction:
-1 -1 -1
0 0 0
1 1 1
musolution.in
• Handcrafted by BackkBenchers community
Scanned by CamScanner
, 11,nag• Segmentation
-----::::~::::~------------------~B~E~l~SE~M~-~7
1
·
AbOVe mask will find edges in ho .
. .
direction.
nzontal direct" ..
ion and it 1s because that zeros column is in horizontal
'
i When you will convolve this
. mask onto an im .
This mask will prominent th h . age It would prominent horizontal edges in the image.
3. e onzontal ed . .
It also works on the principle f ges in an image.
-4. o above mask d
particular edge. an calculates difference among the pixel intensities of a
-
Q2,
------------------------------------------------------------------------------------------------------
Explain image edge detection using Sobel
Ans:
[PI Medium)
JMACE EDGE DETECTION USING SOBEL OPERATOR:
-2 0 2
-1 0 1
1. This mask works ex actly same as the Prew itt operator vertical mask.
2. There is only one d ifference th at is it has "2" and "-2" values in cent er of first and third column.
3. When applied on an image this m ask will highlight the vertical edges.
musolution.in
hers community Page5lof55
• Handcrafted by eackkBenc
Scanned by CamScanner
6
I Image Segmentation
9 · Th.is ·increase the edge intensity and it becomes enhance d co mparatively to the original lrnage.
-2
0
2
l.
Above mask Will find edges in horizontal direction and it is because that zeros column is in hori2:0fltl(
direction.
2. When you will convolve this mask onto an image it would prominent horizontal edges in the irnag._
The only difference between it is that it have 2 and -2 as a center element of first and third rcYW.
3. This mask will prominent the horizontal edges in an image.
4. It also works on the principle of above mask and calculates difference among the pixel intensities Of a
particular edge.
5. As the center row of mask is consist of zeros so it does not include the original values of edge in the
image but rather it calculate the difference of above and below pixel intensities of the particular edge_
6. Thus increasing the sudden change of intensities and making the edge more visible.
------------------------------------------------------------------------------------------------------------------------------
Q3. Explain_ image edge detection using Laplacian Operator
Ans: [P I MedhnJ
IMAGE EDGE DETE·CTION USING LAPl,.l\CIAN OPERATOR~
l. Laµlacian Operator is also a derivative operator which is used to find edges in an image.
2. The major difference between Laplacian and other operators like Prewitt, Sobel, Robinson and Kirsch
is that these all are first order derivative masks but Laplacian is a second order derivative mask.
3. In this mask we have two further classifications one is Pcsitive Laplacian Operator and other Is
Negative Laplacian Operator.
4. Another difference between Laplacian and other operators is that unlike other operators LaplacilJI
didn't take out edges in any particular direction but it take out edges in following classification.
a. Inward Edges
b . Outward Edges
1. In Positive Laplacian we have standard mask 1n which center element of the mask should be negative
and corner elements of mask should be zero.
0 1 0
1 -4 1
0 1 0
musolution.in
• Handcrafted by Bac:kkBenc:hers Community
PageSJ
Scanned by CamScanner
I 1rna9• Segmentation
...--
6
BE ISEM-7
~ Laplacian Operatcm
In negative Laplacian 0
1. Perator w
pasr·t·rve. e also have a standard mask, In which center element should be
0 -l 0
-1 4 -1
0 -1 0
rs. Negative Laplacian operator is use t 0 t k
a e out inward edges in an image
musolution.in
,. . Page54of55
Scanned by CamScanner
BE I SI . .
. =D:_:i_:g.:_:it::::a:l.:::S:lg~n_::a::l~&:._:lm=a~g:e~P:r:o:c:e:ss~i:n~g------------------~----~
· BackkBenchers Community & become the Student Am bassador to represent your college a,
Jorn '•tit
15% Discount.
·····-··················································································································------
Be the Technical Content Writer with BackkBenchers and earn upto l00 Rs. per l0 Marks Questions.
------------------------------------------------------------------·---------·-----...----------------------------------------------
Buy & Sell Final Year Projects with BackkBenchers. Project Charge upto 10,000 .
•••
·······················································································~································---·-··-
Follow us on Social Media Profiles to get notified
·······-··-·--·················································································································--
----------------------------
musolution.in
• Handcrafted by BackkBcrnchers C~mmunity
Pagessotlf
Scanned by CamScanner