A Lecture on
Introduction
to
Image Restoration
Presented By
Kalyan Acharjya
Assistant Professor, Dept. of ECE
Jaipur National University
10/22/2014
Lecture on Image Restoration
2
By Kalyan Acharjya,JNU Jaipur,India
Contact
:[email protected]
10/22/2014
Objective of Two Hour Presentation
To introduce the basic concept of Image
Restoration in Digital Image Processing
10/22/2014
Sorry, shamelessly I opened the lock without prior permission taken
from the original owner. Some images used in this presentation contents
are copied from Books without permission.
Only Original Owner has full rights reserved for copied images.
This PPT is only for fair academic use.
Kalyan Acharjya
10/22/2014
Acknowledgement
The PPT was designed with the help of materials of following authors.
Gonzalez & Woods-DIP Books
Prof. P.K. Biswas, IIT Kharagpur
Dr. ir.Aleksandra Pizurika, Universiteit Hent.
Gleb V. Teheslavski.
Yu Hen Yu.
Zhou Wang, University of Texas.
10/22/2014
Outlines
6
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
Lets start
7
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
What is Image Restoration.
10/22/2014
What is Image Restoration?
9
Image restoration attempts to restore images that have been degraded
Identify the degradation process and attempt to reverse it.
Almost Similar to image enhancement, but more objective.
Fig: Degraded image
Fig: Restored image
10/22/2014
Where we reached?
10
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
Image enhancement vs. Image Restoration
11
Image restoration assumes a degradation model that is known or can be
estimated.
Original content and quality does not mean Good looking or appearance.
Image Enhancement is subjective, where as image restoration is objective
process.
Image restoration try to recover original image from degraded with prior
knowledge of degradation process.
Restoration involves modeling of degradation and applying the inverse
process in order to recover the original image.
Although the restore image is not the original image, its approximation of
actual image.
10/22/2014
Where we reached?
12
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
Degradation Model?
13
Objective: To restore a degraded/distorted image to its original content
and quality.
g(x,y)
f(x,y)
Degradation
Function h
Degradation
Restoration
Filters
(x,y)
f(x,y)
Restoration
Spatial Domain: g(x,y)=h(x,y)*f(x,y)+ (x,y)
Frequency Domain: G(u,v)=H(u,v)F(u,v)+ (u,v)
Matrix: G=HF+
10/22/2014
Going On.!
14
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
Noise Models and Their PDF
15
Different models for the image
noise term (x, y)
Gaussian
Rayleigh
Gaussian
Most common model
Rayleigh
Erlang or Gamma
Exponential
Uniform
Impulse
Erlang
Exponential
Uniform
Impulse
Salt and pepper noise
10/22/2014
Noise Models Effects
16
Histogram to go here
Fig: Original Image
Fig: Original Image histogram
10/22/2014
Noise Models Effects contd1
17
10/22/2014
Noise Models Effects contd2
18
10/22/2014
Going On.!
19
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
Estimation of Degradation Model.
20
Weather the spatial or frequency domain or Matrix, in all cases knowledge
of degradation function is important.
Estimation of H is important in image restoration.
There are mainly three ways to estimate the H as follows-
By Observation
By Experimentation.
Mathematical Modeling
After approximation the degradation function,
we apply the BLIND
CONVOLUTION to restore the original image.
10/22/2014
Observation
21
No knowledge of degraded function is given.
Observing on g(x,y), try to estimate the degraded function in the region
which have simpler structure.
gs(x,y) Gs(u,v)
fs(x,y) Fs(u,v)
Hs(u,v)= Gs(u,v)/ Fs(u,v)
10/22/2014
Observation contd
22
gs(x,y)
Gs(u,v)
Hs(u,v)= Gs(u,v)/ Fs(u,v)
Fs(u,v)
fs(x,y)
10/22/2014
Experimentation
23
Try to imaging set-up similar to original.
Impulse response and impulse simulation.
Objective to find H which have similar result of degradation as original
one.
Impulse
Impulse Response
Fig: Impulse Simulation
10/22/2014
Experimentation contd
24
Here f(x,y) is impulse.
F(u,v)=>A (a constant).
G(u,v)=H(u,v)F(u,v).
H(u,v)=G(u,v)/A.
Objective is training and testing.
Never testing on training data.
Note: The intensity of impulse is very high, otherwise noise can dominate to
impulse.
10/22/2014
Mathematical Modeling
25
If you have the mathematical model, you have inside the degradation
process.
Atmospheric turbulence can be possible to mapping in mathematical model.
One e.g. of mathematical model
k gives the nature of turbulence
10/22/2014
Mathematical Modeling contd..
Atmospheric Turbulence blur examples
26
Fig: Negligible Turbulence
Fig: Severe Turbulence, k=0.0025
Fig: Mid Turbulence, k=0.001
Fig: Low Turbulence, k=0.00025
10/22/2014
Present Position
27
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
Restoration Techniques.
28
Inverse Filtering.
Minimum Mean Squares Errors.
Weiner Filtering.
Constrained Least Square Filter.
Non linear filtering
Advanced Restoration Technique.
10/22/2014
Filter used for Restoration Process
29
Mean filters
Arithmetic mean filter
Geometric mean filter
Harmonic mean filter
Contra-harmonic mean filter
Adaptive filters
Adaptive local noise reduction
filter.
Adaptive median filter
Order statistics filters
Median filter
Max and min filters
Mid-point filter
alpha-trimmed filters
10/22/2014
Filtering to Remove Noise-AMF
30
Use spatial filters of different kinds to remove different kinds of noise
Arithmetic Mean :
f ( x, y ) 1
g ( s, t )
mn ( s ,t )S xy
This is implemented as the simple smoothing filter Blurs the image to
remove noise.
1/
1/
1/
1/
1/
1/
1/
1/
1/
9
10/22/2014
Filtering to Remove Noise-GMF
31
Geometric Mean:
f ( x, y ) g ( s, t )
( s ,t )S xy
1
mn
Achieves similar smoothing to the arithmetic mean, but tends to lose less
image detail.
10/22/2014
Filtering to Remove Noise-HMF
32
Harmonic Mean:
f ( x, y )
mn
( s ,t )S xy
1
g ( s, t )
Works well for salt noise, but fails for pepper noise
Satisfactory result in other kinds of noise such as Gaussian noise
10/22/2014
Filtering to Remove Noise-CHMF
33
Contra-harmonic Mean:
f ( x, y )
g ( s, t )
Q 1
( s ,t )S xy
Q
g
(
s
,
t
)
( s ,t )S xy
Q is the order of the filter and adjusting its value changes the filters
behaviour.
Positive values of
Q eliminate pepper noise.
Negative values of
Q eliminate salt noise.
10/22/2014
Result of AMF and GMF
34
Fig: Original Image
Fig: Gaussian Noise
Fig: Result of 3*3 AM
Fig: Result of 3*3 GM
10/22/2014
Result of Contra-harmonic Mean Filter
35
Fig: Original Image with Pepper noise
Fig: Original Image with Salt noise
Fig: After filter by 3*3 CHF, Q=1.5
Fig: After filter by 3*3 CHF, Q=-1.5
10/22/2014
Beware: Q value in Contra-harmonic Filter
36
Choosing the wrong value for Q when using the contra-harmonic filter can
have drastic results.
10/22/2014
Order Statistics Filters
37
Spatial filters that are based on ordering the pixel values that make up
the neighbourhood operated on by the filter
Useful spatial filters include
Median filter.
Maximum and Minimum filter.
Midpoint filter.
Alpha trimmed mean filter.
10/22/2014
Median Filter
38
Median Filter:
f ( x, y) median{g ( s, t )}
( s ,t )S xy
Excellent at noise removal, without the smoothing effects that can occur
with other smoothing filters
Best result for removing salt and pepper noise.
10/22/2014
Maximum and Minimum Filter
39
Max Filter:
f ( x, y) max {g ( s, t )}
( s ,t )S xy
Min Filter:
f ( x, y) min {g ( s, t )}
( s ,t )S xy
Max filter is good for pepper noise and min is good for salt noise
10/22/2014
Midpoint Filter
40
Midpoint Filter:
f ( x, y) 1 max {g ( s, t )} min {g ( s, t )}
( s ,t )S xy
2 ( s ,t )S xy
Good for random Gaussian and uniform noise
10/22/2014
Alpha-Trimmed Mean Filter
41
Alpha-Trimmed Mean Filter:
f ( x, y )
1
mn d
g ( s, t )
( s ,t )S xy
d/2 lowest and d/2 highest grey levels, so gr(s, t)
represents the remaining mn d pixels
Here deleted the
10/22/2014
Result of Median Filter
42
Fig 1: Salt & Pepper noise
Fig3: Result of 2 pass Med 3*3
Fig2: Result of 1 pass Med 3*3
Fig4: Result of 3 pass Med 3*3
10/22/2014
Result of Max and Min Filter
43
Fig: Corrupted by Pepper Noise
Fig: Filtering Above,3*3 Max Filter
Fig: Corrupted by Salt Noise
Fig: Filtering Above,3*3 Min Filter
10/22/2014
Periodic Noise
44
Typically arises due to electrical or electromagnetic interference.
Gives rise to regular noise patterns in an image
Frequency domain techniques in the Fourier domain are most effective at
removing periodic noise
Fig: periodic Noise
10/22/2014
Band Reject Filters
45
Removing periodic noise form an image involves removing a particular range of
frequencies from that image.
Band reject filters can be used for this purpose.
An ideal band reject filter is given as follows:
1 if D(u, v) D0 2
W
W
H (u, v) 0 if D0 D(u, v) D0
2
2
1 if D(u, v) D0 W
2
10/22/2014
Band Reject Filters contd..
46
The ideal band reject filter is shown below, along with Butterworth
and Gaussian versions of the filter.
Ideal Band
Reject Filter
Butterworth
Band Reject
Filter (of order 1)
Gaussian
Band Reject
Filter
10/22/2014
Result of Band Reject Filter
47
Fig: Corrupted by Sinusoidal Noise
Fig: Fourier spectrum of Corrupted Image
Fig: Butterworth Band Reject Filter
Fig :Filtered image
10/22/2014
Conclusions-What we learnt
48
Restore the original image from degraded image, if u have clue about
degradation function, is called image restoration.
The main objective should be estimate the degradation function.
If you are able to estimate the H, then follow the inverse of degradation
process of an image.
Weather spatial or frequency domain.
Spatial domain techniques are particularly useful for removing random
noise.
Frequency domain techniques are particularly useful for removing periodic
noise.
10/22/2014
Advanced Image Restoration
Adaptive Processing
Spatial adaptive
Frequency adaptive
Nonlinear Processing
Thresholding, coring
Iterative restoration
Advanced Transformation / Modeling
Advanced image transforms, e.g., wavelet
Statistical image modeling
Blind Deblurring or Deconvolution
10/22/2014
For advanced Image Restoration (Adaptive Filtering or Nonlinear Filtering etc.), Please
referred the book of Gonzalez and Woods, Digital Image Processing, Pearson Education
or any other standard Digital Image Processing Books.
OR
OR
10/22/2014
Lets conclude..!
51
What is Image Restoration.
Some Basics Filter
Image Enhancement vs.
Advanced Image Restoration.
Image Restoration.
Image Degradation Model.
Noise Models.
Conclusions.
Tools for DIP.
Applications.
Estimation of Degradation
Model.
Restoration Techniques.
10/22/2014
Conclusions-What we learnt
52
Restore the original image from degraded image, if u have clue about
degradation function is called image restoration.
The main objective should be estimate the degradation function.
If you are able to estimate the H, then follow the inverse of degradation
process of an image.
Weather spatial or frequency domain.
Spatial domain techniques are particularly useful for removing random
noise.
Frequency domain techniques are particularly useful for removing periodic
noise.
10/22/2014
Where you start ?
Digital Image Processing !
53
10/22/2014
Popular Image Processing Software Tools
54
CVIP tools
(Computer Vision and Image Processing tools)
Intel Open Computer Vision Library
Microsoft Vision SDL Library
MATLAB
KHOROS
10/22/2014
Applications of
Digital Image Processing?
55
10/22/2014
Applications of Digital Image Processing
56
Identification.
Computer Vision or Robot vision.
Steganography.
Image Enhancement.
Image Analysis in Medical.
Morphological Image Analysis.
Space Image Analysis.
Bottling and IC Industry.etc.
10/22/2014
[email protected]
57
10/22/2014
https://twitter.com/Kalyan_online
[email protected]
58
10/22/2014
59
10/22/2014
