Spatial Domain Linear Spatial Domain Linear Filtering

Spatial Domain Linear
Filtering
Yao Wang
Polytechnic University, Brooklyn, NY 11201
With contribution from Zhu Liu, Onur Guleryuz, and

Gonzalez/Woods, Digital Image Processing, 2ed
Outline
• Introduction
• Noise remo
removalal using
sing lo
low-pass
pass filters
• Sharpening by edge enhancement
– Edge detection using high-pass filters
– Edge enhancement by high emphasis filters
• Edge detection
– First order gradient
g
– Second order gradient
• Summary
Yao Wang, NYU-Poly EL5123: Spatial Filtering 2
Problems
• Smoothing
– Noise removal
– Detail preserving image smoothing
• Sharpening
Sh i
– Edge enhancement
– Detail focusing
• Edge detection

Approaches
• Spatial domain operation or filtering (the
processed value for the current pixel
depends on both itself and surrounding
pixels)
– Linear filtering
– Non-linear
Non linear filtering
• Rank order filtering including median
• Morphological filtering
– Adaptive filtering

Noise Removal (Image Smoothing)
• An image may be “dirty” (with dots, speckles,stains)
• Noise removal:
– To remove speckles/dots on an image
– Dots can be modeled as impulses (salt-and-pepper or speckle)
or continuously varying (Gaussian noise)
– Can be removed by taking mean or median values of
neighboring pixels (e.g. 3x3 window)
– Equivalent to low
low-pass
pass filtering
• Problem with low-pass filtering
– May blur edges
– More advanced techniques: adaptive, edge preserving

Example

Averaging Filter: An Intuitive Approach
• Replace each pixel by the average of
pixels in a square window surrounding this
pixel
1
g (m, n)  ( f (m  1, n  1)  f (m  1, n)  f (m  1, n  1)
9
 f (m, n  1)  f (m, n)  f (m, n  1)
 f (m  1, n  1)  f (m  1, n)  f (m  1, n  1))
• Trade-off between noise removal and

detail preserving:
– Larger window -> can remove noise more
effectively,
y, but also blur the details/edges
g

Example: 3x3 average
100 100 100 100 100

100 200 205 203 100
100 195 200 200 100
100 200 205 195 100
100 100 100 100 100
100 100 100 100 100

100 144 167 145 100
100 167 200 168 100
100 144 166 144 100
100 100 100 100 100

Example

Weighted Averaging Filter
• Instead of averaging all the pixel values in the
window, give the closer-by
closer by pixels higher
weighting, and far-away pixels lower weighting.
k1 l1
g (m, n)   h(k , l ) f (m  k , n  l )  f (m, n) * h(m, n)
k  k 0 l  l0
• This type of operation is in fact 2-D

2 D linear
convolution of f(m,n) by a filter h(m,n).
• Weighted
g averageg filter retains low frequency
q y
and suppresses high frequency = low-pass filter

Example Weighting Mask
1 1 1 1 2 1
1
 1 1 1 1
 2 4 2
9 16
1 1 1 1 2 1
All weights must sum to one

Example: Weighted Average
100 100 100 100 100
100 200 205 203 100
100 195 200 200 100
100 200 205 195 100
1 1 1 1 2 1
1 100 100 100 100 100 1
 
9 1 1 1 16 2 4 2
1 1 1 1 2 1
100 100 100 100 100 100 100 100 100 100
100 144 167 145 100 100 156 176 158 100
100 167 200 168 100 100 174 201 175 100
100 144 166 144 100 100 156 175 156 100
100 100 100 100 100 100 100 100 100 100

Example
Average filtered image Weighted Average

Original image filtered image

Symmetrical Filter
• The smoothing filter is generally symmetrical in both
horizontal and vertical directions, as well as diagonal.
h(m, n)  h(m, n)  h( m, n)  h(m, n) -W
• Simplified calculation. Hh
k1 l1 k=0
0
g (m, n)   h(k , l ) f (m  k , n  l )  f (m, n) * h(m, n)
k  k 0 l  l0
W
-W l=0 W
k1  k0  l1  l0  W
g(m, n)  h(0,0) f (m, n)
W
 h(0,l)( f (m, n  l)  f (m, n  l)  f (m  l, n)  f (m  l, n))
l 1
W W
 h(k,l)( f (m  k, n  l)  f (m  k, n  l)  f (m  k, n  l)  f (m  k, n  l))
k 1 l 1

Computation Cost
• General (2W+1) by (2W+1) filter
– (2W+1)2 multiplications per processed pixel
W W
g (m, n)    h( k , l ) f ( m  k , n  l )
k  W l  W
• Symmetrical filter
– 1+W+W2 ~=(W+1)
(W+1)2 multiplications
• Saving
– W = 2,
2 and 512x512 image
– Saving 512*512*(52-32)=222
– More savings for larger W

Common Smoothing Filters
1 1 1 1 2 1  1 b 1
1   ; 1  2 4 2 ; b b 2 b , with b  1.
1
1 2 1
10   16   (b  2) 2 
1 1 1 1 2 1  1 b 1
• Are any of these separable?

• Criteria for designing a smoothing filter
– h(k,l)≥0
h(k l)≥0 (function as averaging)
k1 l1
–  h(k , l )  1 (the mean value is preserved)

k  k 0 l  l0

Smoothing Filter is a Low-pass Filter
• The 2D Fourier Transform of a smoothing filter
• Image signal usually varies slowly, and noise is usually a wide band
signal.
• A l i a llow pass filt
Applying filter removes th
the hi
high-frequency
hf partt off the
th
noise.
• Image edges and other details are blurred

Interpretation in Freq Domain
Filter response
Low-passed image spectrum
Original image spectrum
Noise spectrum
0 f

Spectrum of Example Low Pass Filterrs
1 1 1 1 1 b 1 1
H  1 1 1  11 1 1;
1 1 b b 2 b   b 1 b 1;
1 1
H
9 9 ( 2  b) 2  ( 2  b) 2  
1 1 1 1 b 1 1
1
H (u , v)  1  2 cos(2u ) 1  2 cos(2v)  / 9 H (u , v)  b  2 cos(2u ) b  2 cos(2v)  /( 2  b) 2
1
1
0.8
0.5
0.6
0 0.4
0.2
-0.5
30 0
30
25
20
20 25
15 20
20
10 10 15
5 10 10
0 0 5
0 0
b=2

Directional Smoothing
• Smoothing filter tend to blur details
• Challenging problem: detail preserving noise removal!
• Di the i-th directional neighborhood
D3
• Usuallyy 4 or 8 directions
• 3 core steps
– Find σi2, the variance in Di D1
– Find i0, so that σi02 ≤σi2
– Filtering in i01direction
g (m, n) 
N io
 ( k ,l )Dio
f (m  k , n  l )
g (m, n)  f (m, n) * hi 0 (m, n)

p{(m cos  i  n sin  i ) 2 }
hi (m, n)  a exp{(

Noise Removal by Averaging Multiple
Images

Noise Removal by Averaging Multiple
Images
• Assume y = x + n, where n ~ N(0, σ2).
• Observe one image, the noise is n
• Observe N images: yi = x + ni, ii=1
1,…,N,
N where ni ~ N(0,
N(0 σ2),
) and ni
are independent.
N N N
1 1 1
y
N

i 1
yi 
N

i 1
( x  ni )  x 
N
n
i 1
i
y  x  n , where n ~ N (0,  2 / N )
• Averaging over N images reduces the noise variance by 1/N.

Edge Detection
• What is an edge?
• Difference bet
between
een edge and line and
point

Characterization of Edges
Ideal step edge
Real edge has a slope
First order derivative:

Maximum at edge location
Second order derivative:

Zero crossing at edge location

Edge Detection Based on First Order
Derivatives
High-pass filtering across edge
• Edge Low-pass filtering along edge
 1 1 1 
h 
 1  1  1
Non-Edge
Non Edge
No
g Yes
f H |g|>T Edge
What if we don’t know edge direction?

Edge Detection Based on Gradients in
Two Orthogonal Directions
• Combine results from directional edge
detectors in two orthogonal directions and
determine the magnitude and direction of
the edge
edge. Non-Edge
fx
g Edge
Hx fx2+fy2 |g|>T
Hy tan-1(fx/fy) theta
fy

Directional Edge Detector
• High-pass (or band-pass) in one direction (simulating
first order derivative)
• Low pass in the orthogonal direction (smooth noise)
LP BP
• Prewitt edge detector
 1  1  1  1  1 0 1 1
1  1  1  1 
H x   0 0 0    0 1 1 1; H y   1 0 1  1 1 0 1
3 3 3 3
 1 1 1   1   1 0 1 1
• Sobel edge detector
 1  2 1 1   1 0 1 1
1  1  1  1 
H x   0 0 0    0 1 2 1; H y   2 0 2  21 0 1
4 4 4 4
 1 2 1   1   1 0 1 1
The sobel filter provides better smoothing along the edge

Freq. Response of Sobel Filter
Sobel Filter for Horizontal Edges :
 1  2  1  1
H x   0 0    0 1 2 1
1 1
0
4 4
 1 2 1   1 
F
Frequency R
Response (DTFT) :

hx  [-1 0 1]  - H x   e j 2u  e  j 2u  2 j sin 2u 
4
1
4

h y  [1 2 1]  - H y  e j 2v  2  e  j 2v  1  cos 2v 
1 1
2

|Hx(u)| |Hy(v)|
0 1/4 1/2 u 0 1/4 1/2 v
Band-pass
Low-pass
Spectrum of the Sobel Filter
10
5
Hx
0
40
40
20 20
0 0
10
5
Hy
0
40
40
20 20
0 0
L
Low pass along
l th
the edge,
d b
band
d pass cross th
the edge
d

Example of Sobel Edge Detector
Original image Filtered image by Hx Filtered image by Hy

How to set threshold?
• Trial and error
• According to edge magnit
magnitude
de distrib
distribution
tion
– E.g assuming only 5% pixels should be edge
pixels then the threshold should be the 95%
pixels,
percentile of the edge magnitude
– Illustrate on board

gx gy gm
Histogram of gm T=100 T=50 T=20

2500
2000
1500
1000
500
0
0 100 200

Problems of previous approach
• Cannot locate edges precisely
• Ramp edges can lead to many edge pixels
detected depending on the threshold T
– T too high:
g may y not detect weak edges
g
– T too small: detected edges too think, noisy points
falsely detected
• Remedy:
– Detecting local maximum of |g| in the normal direction
of the edge,
edge or try all possible 8 direction in a 3x3
neighbor
– Onlyy consider pixels
p with |g| > T

Edge Detection with Many Directional
Filters
• Instead of using two orthogonal directions,
can design multiple directional filters
– 0, 45, 90, 135
• S
See which
hi h one gives
i th
the hi
highest
h t response
in the normal direction

Edge Detection Based on Second Order
Derivative
• Convolve an image with a filter
corresponding to taking second order
derivative (e.g. Laplacian or LoG operator)
• Locate zero
zero-crossing
crossing in the filtered image

Laplacian Operator
2 f 2 f
  2  2
2
x y
f
f
 f ( x  1, y )  f ( x, y )
x
2 f
 f ( x  1, y )  2 f ( x, y )  f ( x  1, y )
x 2
2 f
 f ( x, y  1)  2 f ( x, y )  f ( x, y  1)
y 2
 2f   f ( x  1, y )  f ( x  1, y )  f ( x, y  1)  f ( x, y  1)  4 f ( x, y )
0 1 0   0  1 0 
1  4 1;  1 4  1;
   
0 1 0  0  1 0 

Fourier Transform of Laplacian Operator
 0 1 0   1  1  1
 1 4  1  1 8  1
   
 0  1 0   1  1  1
6 12
10
4
8
2 6
4
0
30 2
0
20 30
25
10 20
25 20
20 15
15 10
10 10
0 5 5
0
0 0
More isotropic

Isotropic Edge Detector
• Laplacian operator can detect changes in
all directions

Pros and Cons
• Can locate edges more accurately
• Can detect edges in various
ario s direction
• But more prone to noise
• Remedy:
– Smoothing
g before applying
pp y g Laplacian
p

Laplacian of Gaussian (LoG)
• To suppress noise, smooth the signal
using a Gaussian filter first
– F(x,y)* G(x,y)
G
• Then
Th apply
l L
Laplacian
l i filt
filter
– F(x,y)*G(x,y)*L(x,y) = F(x,y)* (L(x,y)*G(x,y))
• Equivalent filter: LoG
– H(x,y)=L(x,y)*G(x,y)

Derivation of LoG Filter
• Analog form
x2  y2

G ( x, y )  e 2 2
x2  y2
 2G  2G x 2  y 2  2 2 
2 2
 G ( x, y ) 
2
  e
2x 2 y 4
• Take samples to create filter mask

– Size
Si off maskk nxn, n>=6,
> 6 odd
dd
– Exn

LoG Filter

Laplacian filtered LOG filtered
Note that each strong g edge

g in the original
g image
g corresponds
p to a thin stripe
p
with high intensity in one side and low intensity in the other side.

How to detect zero crossing?
• For each pixel that has low filtered value,
check a 3x3 neighbor
neighbor, to see whether its
two neighbors in opposite direction have
opposite sign and their difference exceeds
a threshold (Marr-Hildreth edge detection
method)

Example

Summary of Edge Detection Method
• First order gradient based:
– Using edge detectors in two orthogonal directions
• For each direction: low-pass along edge, high-pass across
edge
– Using edge detectors in multiple (>2) directions
– Use threshold or detect local maximum across the
edge
g direction
• Second order gradient based
– Laplacian
p is noise-prone,
p LoG is better
– Detect zero crossing
– Isotropic

More on Edge Detection
• Methods discussed so far generally cannot yield
connected thin edge maps
• Need sophisticated post processing
– Thinning
– Connecting broken lines
• Noise can lead to many false edge points
• Even with many years of research, no perfect
edge
d d detectors
t t exist!
i t!
– Canny edge detector: Gaussian smoothing along
edges, high pass in each possible edge direction
• For more on edge detection, See Gonzalez Sec.
10.1, 10.2

Results using MATLAB “edge( )” function
Sobel,T=0.14 LOG, T=0.0051 canny,T=[0.0313,0.0781]
Sobel,T=0.1 LOG, T=0.01 canny,T=[0.1,0.15]

Image Sharpening
• Sharpening : to enhance line structures or other
details in an image
• Enhanced image = original image + scaled
version of the line structures and edges in the
image
• Line structures and edges can be obtained by
applying a high pass filter on the image
• Combined operation is still a weighted averaging
operation, but some weights can be negative.
• In frequency domain, the filter has the “high-
emphasis” character

Filter response (high emphasis)
sharpened image spectrum
Original image spectrum
0 f

Designing Sharpening Filter Using High
Pass Filters
• The desired image is the original plus an
appropriately scaled high-passed image
• Sharpening filter
f s  f  f h
hs (m, n)   (m, n)  hh (m, n)

high emphasis=allpass+highpass
all pass
high pass
0 f

How to design high pass filter
• High pass filter tries to detect changes
• Difference between a current pixel and its
neighbors
– Some coefficients should be negative
g
• Output should be zero if the image is flat
k1 l1
 h(k , l )  0
k  k 0 l  l0
• More rigorous design would start with desired

frequency response and take inverse FT.

Using Laplacian Operator as Highpass
Filter
 0 1 0   1  1  1
 1 4  1  1 8  1
   
 0  1 0   1  1  1
6 12
10
4
8
2 6
4
0
30 2
0
20 30
25
10 20
25 20
20 15
15 10
10 10
0 5 5
0
0 0
More isotropic

Example
f(x) g(x)=f(x)*hh(x) fs(x)=f(x)+ag(x)
x x x
 0 1 0   0 1 0 
H h   1 4  1  H s   1 8  1 with   1.
1 1
4 4
 0  1 0   0  1 0 
 1  1  1  1  1  1
H h   1 8  1  H s   1 16  1 with   1.
1 1
8 8
 1  1  1  1  1  1

Example of Sharpening Using Laplacian
Operator
 0 1 0 
H h   1 8  1
1
4 
 0  1 0 

Example of Sharpening Using Laplacian
Operator
4
 8

Designing Sharpening Filtering Using Low
Pass Filters ((Unsharp
p Masking)
g)
• The sharpened = original – a*lowpass
fs 
1
 f  f l , 0    1
1 
• Sharpening filter
hs (m, n) 
1
 (m, n)  hl (m, n) 
1 

high emphasis=allpass- a*lowpass
all pass
low pass
0 f

Example
f(x) fl(x)=f(x)*hl(x) fs(x)=f(x)- fl(x)
x x x
1 2 1    1  2  1
1   1   1
H l   2 4 2  H s    2 60  2 with   .
16 48 4
1 2 1    1  2  1 
Criteria for designing sharpening filters: There are both positive and negative
Coefficients h(0
Coefficients, h(0,0)
0) > 0
0, and ∑h(k,
∑h(k l)=1
l) 1 (to keep the same mean value)
value).

Example
Original image Sharpened image

Challenges of Noise Removal and Image
Sharpening
p g
• How to smooth the noise without blurring
the details too much?
• How to enhance edges without amplifying
noise?
• Still a active research area
– Wavelet domain processing

Wavelet-Domain Filtering
Courtesy of Ivan Selesnick

Feature Enhancement by
Subtraction
Taking an image without injecting a contrast agent first. Then take the image again after
the organ is injected some special contrast agent (which go into the bloodstreams only).
Then subtract the two images --- A popular technique in medical imaging

Summary
• Noise removal using low-pass filters
– Averaging and weighted averaging filter
• Edge detection
– Using isotropic high-pass filters: eg. 2D Laplacian
– Using two orthogonal high-pass filters: eg. Sobel
– Using many directional high-pass filters
• Image sharpening by high emphasis filters
– High pass filtering using Laplacian operators
– Sharpening using high pass filter (low pass+ a* high pass)
– Sharpening using low pass filter (unsharp masking)(original-
a*lowpass)
p )
• Given a filter, you should be able to tell what does it do
(smoothing, edge detection, sharpening?) by looking at
its coefficients
coefficients, and also through DTFT

Reading
• R. Gonzalez, “Digital Image Processing”,
Section 3
3.4
4~33.8,
8 10
10.1,
1 10.2
10 2
– Note that in Gonzalez book: Edge detection
filtering are referred to sharpening
sharpening.
• A. K. Jain, “Fundamentals of Digital Image
Processing ” Section 7
Processing, 7.4
4 (except median
filter).

Written Assignment
1. For each filter given below, compute its Fourier transform, and illustrate its magnitude
response. Determine what is its function (smoothing, edge enhancement or edge detection?)
based on the filter coefficients as well as its frequency response. For each filter, determine
whether it is separable? If yes,
yes compute the FT separately and explain the function of each
1D filter. If not, compute the FT directly.
1 2 1   1  1  1  0 1 0 
H 1  2 4 2 H 2   1 8  1;
1
H 3   1 5  1
16
1 2 1   1  1  1  0  1 0 

Written Assignment (cnt’d)
2. Recall that the Laplacian of Gaussian filter is the Laplacian of a Gaussian function. Let the
Gaussian function be given by 
x2  y2
G ( x, y )  e 2 2
show that the LOG filter can be written as

x2  y2
x 2  y 2  2 2 
L( x , y )  e 2 2
4
3
3. Define the a discrete Laplacian of Gaussian (LoG) filter of size 7x7 by sampling the above
continuous LoG function, assuming =1. Determine its discrete space Fourier transform.
Illustrate its magnitude response. Explain its function. You can use the “freqz2( )” function to
determine and illustrate its DTFT.

Computer Assignment
1. Write a Matlab or C-program to simulate noise removal. First create a noisy image, by adding
zero mean Gaussian random noise to your image (In matlab, the noise can be generated using
“imnoise()”). You can specify the noise variance in “imnoise( )”). Then apply an averaging filter
to the noise added image. For a chosen variance of the added noise, you need to try different
window sizes (from 3x3 to 7x7) to see which one gives you the best trade-off between noise
removal and blurring. Hand in your program, the original noise-added images at two different
noise levels (0.01 and 0.1) and the corresponding filtered images with the best window sizes.
Write down your observation. For the filtering operation, don’t use the matlab “conv2()”
function. Rather, write your own function for the averaging operation. Your program should
allow the user to specify the window size as an input parameter.
2. Write a Matlab or C-program for implementing the following simple edge detection algorithm:
For every pixel: i) find the horizontal and vertical gradients, gx and gy, using the Sobel operator;
ii) calculate the magnitude of the gradient using g m  g x  g y ; and iii) For a chosen
2 2
threshold T, consider the pixel to be an edge pixel if gm ≥ T. Save the resulting edge map (Use
a gray level of 255 to indicate an edge pixel, and a gray level of 0 for a non-edge pixel). Apply
this program to your test image, and observe the resulting edge map with different T values,
until
u t you find
d a T tthat
at gives
g es you a good result.
esu t Hand
a d in you
your p
program
og a aand
d tthe
e edge maps
aps
generated by two different threshold values. Write down your observation. Hint: one automatic
way to determine T is by sorting the pixels based on the gradient magnitudes, and choose T
such that a certain percentage (say 5%) of pixels will be edge pixels. You can use either the
matlab conv2( ) function or write your own code for the filtering part.

3. Using the matlab “edge( )” function, evaluate the edge detection results using the
following methods on an image: “sobel”, “log”,and “canny”. For each method, first use
the default threshold value. Then try to use a threshold that is higher and lower, to
see the effect. Comments on the advantages and disadvantages of each method.
For problems 2 and 3, try to use an image that has a lot of edges. A good set of test
images can be found under the matlab directory …./toolbox/images/imdemos

Spatial Domain Linear Spatial Domain Linear Filtering

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

Spatial Domain Linear Spatial Domain Linear Filtering

Uploaded by

Copyright:

Available Formats

Spatial Domain Linear

With contribution from Zhu Liu, Onur Guleryuz, and

Yao Wang, NYU-Poly EL5123: Spatial Filtering 3

Yao Wang, NYU-Poly EL5123: Spatial Filtering 4

Yao Wang, NYU-Poly EL5123: Spatial Filtering 5

Yao Wang, NYU-Poly EL5123: Spatial Filtering 6

• Trade-off between noise removal and

Yao Wang, NYU-Poly EL5123: Spatial Filtering 7

100 100 100 100 100

100 100 100 100 100

Yao Wang, NYU-Poly EL5123: Spatial Filtering 8

Yao Wang, NYU-Poly EL5123: Spatial Filtering 9

• This type of operation is in fact 2-D

Yao Wang, NYU-Poly EL5123: Spatial Filtering 10

All weights must sum to one

Yao Wang, NYU-Poly EL5123: Spatial Filtering 11

Yao Wang, NYU-Poly EL5123: Spatial Filtering 12

Average filtered image Weighted Average

Yao Wang, NYU-Poly EL5123: Spatial Filtering 13

Yao Wang, NYU-Poly EL5123: Spatial Filtering 14

Yao Wang, NYU-Poly EL5123: Spatial Filtering 15

• Are any of these separable?

–  h(k , l )  1 (the mean value is preserved)

Yao Wang, NYU-Poly EL5123: Spatial Filtering 16

Yao Wang, NYU-Poly EL5123: Spatial Filtering 17

Low-passed image spectrum

Original image spectrum

Yao Wang, NYU-Poly EL5123: Spatial Filtering 18

Yao Wang, NYU-Poly EL5123: Spatial Filtering 19

• Di the i-th directional neighborhood

g (m, n)  f (m, n) * hi 0 (m, n)

Yao Wang, NYU-Poly EL5123: Spatial Filtering 20

Yao Wang, NYU-Poly EL5123: Spatial Filtering 21

• Averaging over N images reduces the noise variance by 1/N.

Yao Wang, NYU-Poly EL5123: Spatial Filtering 22

Yao Wang, NYU-Poly EL5123: Spatial Filtering 23

Real edge has a slope

First order derivative:

Second order derivative:

Yao Wang, NYU-Poly EL5123: Spatial Filtering 24

What if we don’t know edge direction?

Yao Wang, NYU-Poly EL5123: Spatial Filtering 25

Yao Wang, NYU-Poly EL5123: Spatial Filtering 26

The sobel filter provides better smoothing along the edge

Yao Wang, NYU-Poly EL5123: Spatial Filtering 27

0 1/4 1/2 u 0 1/4 1/2 v

Yao Wang, NYU-Poly EL5123: Spatial Filtering 29

Original image Filtered image by Hx Filtered image by Hy

Yao Wang, NYU-Poly EL5123: Spatial Filtering 30

Yao Wang, NYU-Poly EL5123: Spatial Filtering 31

Histogram of gm T=100 T=50 T=20

Yao Wang, NYU-Poly EL5123: Spatial Filtering 32

Yao Wang, NYU-Poly EL5123: Spatial Filtering 33

Yao Wang, NYU-Poly EL5123: Spatial Filtering 34

Yao Wang, NYU-Poly EL5123: Spatial Filtering 35

Yao Wang, NYU-Poly EL5123: Spatial Filtering 36

Yao Wang, NYU-Poly EL5123: Spatial Filtering 37

Yao Wang, NYU-Poly EL5123: Spatial Filtering 38

Yao Wang, NYU-Poly EL5123: Spatial Filtering 39

Yao Wang, NYU-Poly EL5123: Spatial Filtering 40

• Take samples to create filter mask

Yao Wang, NYU-Poly EL5123: Spatial Filtering 41

Yao Wang, NYU-Poly EL5123: Spatial Filtering 42