MULTIRESOLUTION
ANALYSIS
Image Representation
• Wavelet Pyramids
•Gaussian pyramids
• Laplacian Pyramids
• Applications
What is a good representation
for image analysis?
• Pixel domain representation tells you
“where” (pixel location), but not
“what”.
– In space, this representation is too localized
• Fourier transform domain tells you
“what” (textural properties), but not
“where”.
– In space, this representation is too spread out.
• Want an image representation that gives
you a local description of image events
—what is happening where.
– That representation might be “just right”.
Image representation
Fourier Analysis
Breaks down a signal into constituent
sinusoids of different frequencies
In other words: Transform the view of the
signal from time-base to frequency-base.
What’s wrong with Fourier?
By using Fourier Transform , we loose
the time information : WHEN did a
particular event take place ?
FT can not locate drift, trends, abrupt
changes, beginning and ends of
events, etc.
Calculating use complex numbers.
Time and Space
definition
Time – for one dimension •
waves we start point shifting
from source to end in time
. scale
Space – for image point •
. shifting is two dimensional
. Here they are synonyms •
Space-Frequency Tiling
Freq.
Standard basis
Spatial
Freq.
Fourier basis
Spatial
Freq.
Wavelet basis
Spatial
Short Time Fourier Analysis
In order to analyze small section of a
signal, Denis Gabor (1946), developed a
technique, based on the FT and using
windowing : STFT
STFT (or: Gabor Transform)
A compromise between time-based
and frequency-based views of a signal.
both time and frequency are
represented in limited precision.
The precision is determined by the size
of the window.
Once you choose a particular size for
the time window - it will be the same
for all frequencies.
What’s wrong with Gabor?
Many signals require a more flexible
approach - so we can vary the window
size to determine more accurately
either time or frequency.
Space-Frequency Tiling
Freq.
Standard basis
Spatial
Freq.
Fourier basis
Spatial
Freq.
Wavelet basis
Spatial
Wavelet's properties
Short time localized waves with •
.zero integral value
.Possibility of time shifting •
.Flexibility •
Scaling
Wavelet analysis produces a time-
scale view of the signal.
Scaling means stretching or
compressing of the signal.
scale factor (a) for sine waves:
f ( t ) sin( t ) ; a 1
f ( t ) sin( 2t ) ; a 1 2
f ( t ) sin( 4t ) ; a 1 4
Scaling (Cont’d)
Scale factor works exactly the same
with wavelets:
f ( t ) (t ) ; a 1
f ( t ) ( 2 t ) ; a 1 2
f ( t ) ( 4 t ) ; a 1 4
Wavelet function
b – shift •
a , b x 1
a
x b
a
coefficient
a – scale •
coefficient
a , bx , by x , y 1a
x bx
a , y by
a 2D function •
a , bx , by x , y
a , bx , by x , y
1
1a x bx
x abx ,
,
y by
y aby
a a a
Various Wavelet basis
Wavelet - Frequency domain
Wavelet bands are split recursively
image
H L
H L
H L
Discrete Wavelet Transform
D DWT for Image-2 •
Discrete Wavelet Transform
Discrete Wavelet Transform
D DWT for Image-2 •
Wavelet
Decomposition:Example
LENA
LH
HL HH
Wavelet - Frequency domain
Wavelet decomposition - 2D
Frequency domain
Horizontal high pass Vertical high pass
Horizontal low pass Vertical low pass
Wavelet - Frequency domain
Apply the wavelet transform separably in both dimensions
Horizontal high pass, Horizontal high pass,
vertical high pass vertical low-pass
Horizontal low pass, Horizontal low pass,
vertical high-pass Vertical low-pass
JPEG2000
v.s. JPEG
low bit-rate performance
Image Pyramids
Image features at different resolutions require filters at
different scales.
Edges (derivatives):
f(x)
f (x)
Space Required for Pyramids
Image Pyramids
Image Pyramid = Hierarchical representation of an image
No details in image -
Low (blurred image)
Resolution low frequencies
High Details in image -
Resolution low+high frequencies
A collection of images at different resolutions.
Image Pyramid
Low resolution
High resolution
Image Pyramid
Frequency Domain
Low resolution
High resolution
Image Blurring = low pass filtering
*
=
~
=
*
=
*
=
Image Pyramid
Low resolution
High resolution
Gaussian pyramid construction
filter mask
Repeat
• Filter
• Subsample
Until minimum resolution reached
• can specify desired number of levels (e.g., 3-level pyramid)
The whole pyramid is only 4/3 the size of the original image!
Gaussian Pyramid
Level n
1X1
Level 1
2n-1 X 2n-1
Level 0
2n X 2n
Image Pyramids
Known as a Gaussian Pyramid [Burt and Adelson, 1983]
• In computer graphics, a mip map [Williams, 1983]
• A precursor to wavelet transform
The Gaussian Pyramid
Low resolution G4 (G3 * gaussian) 2
blur ) 2 down-sample
G3 (G2 * gaussian
down-sam
blur ple
G2 (G1 * gaussian) 2
dow
n-sa
mple
blur
G1 (G0 * gaussian) 2
do
wn
-sa
plem
G0 Image
blur
High resolution
Gaussian Pyramid
Image pyramids
• Gaussian Pyramids
• Laplacian Pyramids
What does blurring take
?away
smoothed (5x5 Gaussian)
High-Pass filter
smoothed – original
Band-pass filtering
Gaussian Pyramid (low-pass images)
Laplacian Pyramid
Need this!
Original
image
How can we reconstruct (collapse) this •
?pyramid into the original image
Laplacian Pyramid
Motivation = Compression, redundancy removal.
compression rates are higher for predictable values.
e.g. values around 0.
G0, G1, .... = the levels of a Gaussian Pyramid.
Predict level Gl from level Gl+1 by Expanding Gl+1 to G’l
Gl+1
Expand
Reduce
Gl G’l
Denote by Ll the error in prediction:
Ll = Gl - G’l
L0, L1, .... = the levels of a Laplacian Pyramid.
?What does blurring take away
original
?What does blurring take away
smoothed (5x5 Gaussian)
?What does blurring take away
smoothed – original
Laplacian Pyramid
Gaussian Laplacian
Pyramid Pyramid
expa
nd
ex
p
- =
an
d
ex
p
- =
an
d
- =
The Laplacian Pyramid
Li Gi expand( Gi 1 )
Gaussian Pyramid Gi Li expand( Gi 1 ) Laplacian Pyramid
Gn Ln Gn
ex pa
nd
G2
an-
L2
exp
d
=
G1 L1
ex
pa -
nd =
G0 L0
- =
Image Pyramids - Comparison
Image pyramid levels = Filter then sample.
Filters:
Gaussian Pyramid
Laplacian Pyramid
Wavelet Pyramid
Gaussian
Frequency
Laplacian
Pyramid Domain
Pyramid
Laplace Pyramid -
No scaling
50
100
50
150
100
200
150
50
250
50 100 150 200 250
200
100
250
150
50 100 150 200 250
200
250
50 100 150 200 250
50
100
50
150
100
200
150
250
50 100 150 200 250
200
250
50 100 150 200 250
from: B.Freeman
Reconstruction of the original image
from the Laplacian Pyramid
Laplacian Gl = Ll + G’l
Pyramid
expand
+ =
expand
+ =
expand
+ =
Original
= Image
Multiresolution Spline - Example
Original - Left Original - Right
Glued Splined
laplacian level 4
laplacian level 2
laplacian level 0
left pyramid right pyramid blended pyramid
Multiresolution Spline
Image Blending
Blending
Multiresolution Spline
When splining two images, transition from one image to
the other should behave:
High Frequencies
Middle Frequencies
Low Frequencies
Multiresolution Spline
High Frequencies
Middle Frequencies
Low Frequencies
Multiresolution Spline - Example
Left Image Right Image
Left + Right Narrow Transition
Wide Transition
(Burt & Adelson)
Multiresolustion Spline - Using Laplacian Pyramid
Multiresolution Spline - Example
Left Image Right Image
Left + Right Narrow Transition
Wide Transition Multiresolution Spline
(Burt & Adelson)
Multiresolution Spline - Example
The colored version
© prof. dmartin
Pyramids in Frequency
Domain
Gaussian Pyramid Laplacian Pyramid
Wavelet Decomposition
Fourier Space
Image Fusion
Multi-scale Transform (MST) = Obtain Pyramid from Image
Inverse Multi-scale Transform (IMST) = Obtain Image from Pyramid