Image Transforms and Image

Enhancement in Frequency Domain

Lecture 5, Feb 25th, 2008

Lexing Xie

EE4830 Digital Image Processing

http://www.ee.columbia.edu/~xlx/ee4830/
thanks to G&W website, Mani Thomas, Min Wu and Wade Trappe for slide materials
 HW clarification
 HW#2 problem 1
 Show: f - r2 f ¼ A f – B blur(f)
 A and B are constants that do not matter, it is up to you
to find appropriate values of A and B, as well as the
appropriate version of the blur function.

 Recap for lecture 4

 roadmap

 2D-DFT definitions and intuitions

 DFT properties, applications
 pros and cons
 DCT
 the return of DFT
 Fourier transform: a continuous
signal can be represented as a
(countable) weighted sum of
sinusoids.
 warm-up brainstorm
 Why do we need image transform?
 why transform?
 Better image processing
 Take into account long-range correlations in space
 Conceptual insights in spatial-frequency information.
what it means to be “smooth, moderate change, fast change, …”
 Fast computation: convolution vs. multiplication

 Alternative representation and sensing

 Obtain transformed data as measurement in radiology images
(medical and astrophysics), inverse transform to recover image

 Efficient storage and transmission

?
 Energy compaction
 Pick a few “representatives” (basis)
 Just store/send the “contribution” from each basis
 outline
 why transform
 2D Fourier transform
 a picture book for DFT and 2D-DFT
 properties
 implementation
 applications
 discrete cosine transform (DCT)
 definition & visualization
 Implementation

next lecture: transform of all flavors, unitary

transform, KLT, others …
 1-D continuous FT
 1D – FT

real(g(x)) imag(g(x))
 1D – DFT of length N
 =0

 =7

x x
 1-D DFT in as basis expansion

Forward transform
real(A) imag(A)

u=0
Inverse transform

basis

u=7

n n
1-D DFT in matrix notations

real(A) imag(A)

u=0

N=8

u=7

n n
 1-D DFT of different lengths

real(A) imag(A)
n
u

N=32
N=8

N=16 N=64
 performing 1D DFT

real-valued input

Note: the coefficients in x and y on this slide are only meant for illustration purposes, which are not numerically accurate.
 another illustration of 1D-DFT

real-valued input

Note: the coefficients in x and y are not numerically accurate

 from 1D to 2D
1D 2D

?
 Computing 2D-DFT

DFT

IDFT

 Discrete, 2-D Fourier & inverse Fourier transforms are implemented

in fft2 and ifft2, respectively
 fftshift: Move origin (DC component) to image center for display
 Example:
>> I = imread(‘test.png’); % Load grayscale image
>> F = fftshift(fft2(I)); % Shifted transform
>> imshow(log(abs(F)),[]); % Show log magnitude
>> imshow(angle(F),[]); % Show phase angle
 2-D Fourier basis

real imag

real( ) imag( )
 2-D FT illustrated
real-valued

real

imag
 notes about 2D-DFT
 Output of the Fourier transform is a complex number
 Decompose the complex number as the magnitude and phase
components
 In Matlab: u = real(z), v = imag(z), r = abs(z), and
theta = angle(z)
 Explaining 2D-DFT

fft2

ifft2
circular convolution and zero padding
zero padded filter and response
zero padded filter and response
observation 1: compacting energy
 observation 2: amplitude vs. phase

 Amplitude: relative prominence of sinusoids

 Phase: relative displacement of sinusoids
 another example: amplitude vs. phase

A = “Aron” P = “Phyllis”
FA = fft2(A) FP = fft2(P)

log(abs(FA)) log(abs(FP))

angle(FA) angle(FP)

ifft2(abs(FA), angle(FP)) ifft2(abs(FP), angle(FA))

Adpated from http://robotics.eecs.berkeley.edu/~sastry/ee20/vision2/vision2.html

 fast implementation of 2-D DFT
 2 Dimensional DFT is separable

1-D DFT
of f(m,n)
w.r.t n

1-D DFT
of F(m,v)
w.r.t m

 1D FFT: O(N¢log2N)
 2D DFT naïve implementation: O(N4)
 2D DFT as 1D FFT for each row and then for
each column
 Implement IDFT as DFT

DFT2

IDFT2
Properties of 2D-DFT
duality result
 outline
 why transform
 2D Fourier transform
 a picture book for DFT and 2D-DFT
 properties
 implementation
 applications
 discrete cosine transform (DCT)
 definition & visualization
 implementation
DFT application #1: fast Convolution

? ?
O(N2)

Spatial filtering
f(x.y)*h(x.y)

?
DFT application #1: fast convolution

O(N2¢log2N) O(N2) O(N2¢log2N)

Spatial filtering
f(x.y)*h(x.y)

O(N4)
 DFT application #2: feature correlation
 Find letter “a” in the following image

bw = imread('text.png'); a = imread(‘letter_a.png');
% Convolution is equivalent to correlation if you rotate the
convolution kernel by 180deg
C = real(ifft2(fft2(bw) .*fft2(rot90(a,2),256,256)));

% Use a threshold that's a little less than max.

% Display showing pixels over threshold.
thresh = .9*max(C(:)); figure, imshow(C > thresh)

from Matlab image processing demos.

 DFT application #3: image filters
 Zoology of image filters
 Smoothing / Sharpening / Others
 Support in time vs. support in frequency
c.f. “FIR / IIR”
 Definition: spatial domain/frequency domain
 Separable / Non-separable
smoothing filters: ideal low-pass
butterworth filters
Gaussian filters
low-pass filter examples
 smoothing filter application 1

text enhancement
 smoothing filter application 2
beautify a photo
high-pass filters
 sobel operator in frequency domain

Question:
Sobel vs. other high-pass
filters?
Spatial vs frequency
domain implementation?
high-pass filter examples
 outline
 why transform
 2D Fourier transform
 a picture book for DFT and 2D-DFT
 properties
 implementation
 applications in enhancement, correlation
 discrete cosine transform (DCT)
 definition & visualization
 implementation
 Is DFT a Good (enough) Transform?
 Theory

 Implementation

 Application
 The Desirables for Image Transforms
 Theory DFT ???
 Inverse transform available X
 Energy conservation (Parsevell) X
 Good for compacting energy ?
 Orthonormal, complete basis
 (sort of) shift- and rotation invariant
X
 Implementation X
 Real-valued x
 Separable X
 Fast to compute w. butterfly-like structure X
 Same implementation for forward and
inverse transform X
 Application
 Useful for image enhancement X
 Capture perceptually meaningful structures X
in images
 DFT vs. DCT
1D-DCT 1D-DFT

a real(a) imag(a)

u=0 u=0

u=7 u=7

n=7
 1-D Discrete Cosine Transform (DCT)
 N 1
  ( 2n  1) k 


Z ( k )  
n 0
z ( n )   ( k ) cos 
 2N 


 z ( n) 
N 1
  ( 2n  1) k 



k 0
Z ( k )   ( k ) cos 
 2N 

1 2
 (0)  ,  (k ) 
N N
 Transform matrix A
 a(k,n) = (0) for k=0

 a(k,n) = (k) cos[(2n+1)/2N] for k>0

 A is real and orthogonal

 rows of A form orthonormal basis
 A is not symmetric!
 DCT is not the real part of unitary DFT!
 1-D DCT
1.0 1.0 100 100

0.0 0.0 0 0

-1.0 -1.0 -100 u=0 -100 u=0 to 4

z(n)
1.0 1.0 100 100

n 0.0 0.0 0 0

Original signal -1.0 -1.0 -100 u=0 to 1 -100 u=0 to 5

1.0 1.0 100 100

0.0 0.0 0 0

-1.0 -1.0 -100 u=0 to 2 -100 u=0 to 6

Z(k)
1.0 1.0 100 100

k 0.0 0.0 0 0

-1.0 -1.0 -100 u=0 to 3 -100 u=0 to 7

Transform coeff.
Basis vectors Reconstructions
 DFT and DCT in Matrix Notations
Matrix notation for 1D transform

1D-DCT 1D-DFT

N=32 A real(A) imag(A)

 From 1D-DCT to 2D-DCT

u=0

u=7

n=7

 Rows of A form a set of orthonormal basis

 A is not symmetric!
 DCT is not the real part of unitary DFT!
basis images: DFT (real) vs DCT
Periodicity Implied by DFT and DCT
 DFT and DCT on Lena
DFT2 DCT2

Shift low-freq
to the center

Assume periodic and zero-padded … Assume reflection …

 Using FFT to implement fast DCT
 Reorder odd and even elements
~z ( n)  z ( 2n ) N
~ for 0  n  1
 z ( N  n  1)  z ( 2n  1) 2

 Split the DCT sum into odd and even terms

 N / 21   ( 4n  1) k  N / 2 1   ( 4n  3) k  
Z ( k )   ( k )   z ( 2n)  cos  
 z ( 2 n  1)  cos  
 n 0  2N  n 0  2N 
 N / 2 1 ~   ( 4n  1) k  N / 2 1 ~   ( 4n  3) k  
  ( k )   z ( n)  cos    z ( N  n  1)  cos   
 n 0  2N n 0 2N 
 N / 2 1 ~   ( 4n  1) k  N 1
  ( 4 N  4n'1) k  
  ( k )   z ( n)  cos  
 ~
z ( n ' )  cos   
 n 0  2 N n ' N / 2 2 N 
N 1
  ( 4n  1) k   N 1
 j 2nk / N 
  ( k ) z ( n)  cos
~
  Re   ( k ) e  jk / 2 N
 ~
z ( n )  e 
n 0  2N   n 0 

 Re  ( k )e  jk / 2 N DFT  ~ z ( n) N 
 The Desirables for Image Transforms
 Theory DFT DCT ???
 Inverse transform available X X
 Energy conservation (Parsevell) X X
 Good for compacting energy ? ?
 Orthonormal, complete basis
 (sort of) shift- and rotation invariant
X X
 Implementation X X
 Real-valued x X
 Separable X X
 Fast to compute w. butterfly-like structure X X
 Same implementation for forward and
inverse transform X X
 Application
 Useful for image enhancement X
 Capture perceptually meaningful structures X
in images
 Summary of Lecture 5
 Why we need image transform
 DFT revisited
 Definitions, properties, observations, implementations, applications
 What do we need for a transform
 DCT

 Coming in Lecture 6:
 Unitary transforms, KL transform, DCT
 examples and optimality for DCT and KLT, other transform flavors,
Wavelets, Applications
 Readings: G&W chapter 4, chapter 5 of Jain has been posted on
Courseworks

 “Transforms” that do not belong to lectures 5-6:

Rodon transform, Hough transform, …