Fourier Optics
Ivan Bazarov
Cornell Physics Department / CLASSE
Outline
• 2D Fourier Transform
• 4-f System
• Examples of spatial frequency filters
• Phase contrast imaging
• Matlab FFT
1 Fourier Optics P3330 Exp Optics FA’2016
Properties of 2D Fourier Transforms
Definitions:
ZZ
i2⇡(xfx +yfy )
F (fx , fy ) = f (x, y)e dxdy
ZZ
f (x, y) = F (fx , fy )ei2⇡(xfx +yfy ) dfx dfy
Linearity:
↵f (x, y) + g(x, y) ! ↵F (fx , fy ) + G(fx , fy )
Scaling:
⇣x y⌘
f , ! |ab|F (afx , bfy )
a b
Shift:
i2⇡(x0 fx +y0 fy )
f (x x0 , y y0 ) ! F (fx , fy )e
2 Fourier Optics P3330 Exp Optics FA’2016
Properties of 2D Fourier Transforms (contd.)
Rotation:
R✓ {f (x, y)} ! R✓ {F (fx , fy )}
Convolution:
ZZ
f (x̃, ỹ)g(x x̃, y ỹ)dx̃dỹ ! F (fx , fy )G(fx , fy )
Parseval’s theorem:
ZZ ZZ
|f (x, y)|2 dxdy = |F (fx , fy )|2 dfx dfy
Slice theorem:
Z
f (x, y)dy ! F (fx , 0)
3 Fourier Optics P3330 Exp Optics FA’2016
4-f System
Collimate Mask FT Filter IFT Out
FFT FFT FFT FFT
Prepare input 1 “4-f system”
Input mask f (x, y) ✓
x0 y0
◆ ✓
x0 y0
◆
Filter mask F , G ,
✓ ◆ FF T FF T FF T FF T
x0 y0 ZZ
Take FT F ,
FF T FF T Inverse FT f (x̃, ỹ)g(x x̃, y ỹ)dx̃dỹ
2D object f(x,y) has been filtered with 2D filter with impulse response g(x,y)
4 Fourier Optics P3330 Exp Optics FA’2016
Coordinate system after the lens: spatial
frequency converter
Thus, the spatial frequency fx is related to coordinate x’ by scaling factor Fl
5 Fourier Optics P3330 Exp Optics FA’2016
–Fourier optics
Simple
Simple Fourier
optical Transform
Fourier transforms
Focal length F = 100 mm
Laser wavelength λ0 = 632 nm
Amplitude cosine, aka diffraction grating
100,000 × 0.632
x′ =
200
λ x = 200 µm
= 316 µm
Rotate object by 45o
x′ = y ′
100,000 × 0.632
=
282.8
= 223 µm
λx = λ y
= 2 × 200 µm
= 282.8 µm
6 Fourier Optics 49P3330 Exp Optics FA’2016
Low pass filter (sharp cutoff)
•Lecture 4 ECE 4606 Undergraduate Optics Lab
–Example
Low pass, sharp cutoff
REAL SPACE FOURIER SPACE
Multiplied by
252.8 µm pinhole
Filter cutoff frequency = 1/500 µm-1
Filter cutoff position = 126.4 µm
Focal length = 100 mm
Laser wavelength = 632 nm
All plots show amplitude of E
Smoothed, but Gibbs ringing
due to sharp filter edges =
7 Robert R. McLeod, University of Colorado Fourier Optics 50 P3330 Exp Optics FA’2016
Low pass filter (smooth cutoff)
•Lecture 4
–Example
ECE 4606 Undergraduate Optics Lab
Low pass, smooth cutoff
REAL SPACE FOURIER SPACE
Multiplied by
Filter cutoff frequency = 1/500 µm-1
Filter cutoff position = 126.4 µm
Edge smoothing = 132 µm
Focal length = 100 mm
Laser wavelength = 632 nm
Now just nicely smoothed
=
8 51
Robert R. McLeod, University of Colorado
Fourier Optics P3330 Exp Optics FA’2016
•Lecture 4
–Example High pass (narrow band) ECE 4606 Undergraduate Optics Lab
High pass, narrowband
REAL SPACE FOURIER SPACE
Multiplied by
105.2 µm “dot”
Filter cutoff frequency = 1/1200 µm-1
Filter cutoff position = 52.6 µm
Edge smoothing = 58.2 µm
Focal length = 100 mm
Laser wavelength = 632 nm
Note sharp edges, darkening of
large, uniform areas (~DC) =
Sharp filter used for clarity
9 Robert R. McLeod, University of Colorado Fourier Optics 52 P3330 Exp Optics FA’2016
•Lecture 4
–Example High pass (high band) ECE 4606 Undergraduate Optics Lab
High pass, wideband
REAL SPACE FOURIER SPACE
Multiplied by
Filter cutoff frequency = 1/300 µm-1
Filter cutoff position = 210.6 µm
Edge smoothing = 46.6 µm
Focal length = 100 mm
Laser wavelength = 632 nm
Only edges remain. Almost a
“line drawing” =
10 Robert R. McLeod, University of Colorado Fourier Optics 53 P3330 Exp Optics FA’2016
Vertical low pass (“smeers” vertically)
•Lecture 4 ECE 4606 Undergraduate Optics Lab
–Example
Vertical low pass
REAL SPACE FOURIER SPACE
Multiplied by
632 µm horiz. slit
Filter cutoff frequency = 1/200 µm-1
Filter cutoff position = 316 µm
Edge smoothing = 93.6 µm
Focal length = 100 mm
Laser wavelength = 632 nm
Horizontal lines at edges of eyes gone
Vertical lines above nose remain.
=
11 Robert R. McLeod, University of Colorado Fourier Optics 54 P3330 Exp Optics FA’2016
Horizontal low pass (“smeers” horizontally)
•Lecture 4 ECE 4606 Undergraduate Optics Lab
–Example
Horizontal low pass
REAL SPACE FOURIER SPACE
Multiplied by
Filter cutoff frequency = 1/200 µm-1
Filter cutoff position = 316 µm 632
Edge smoothing = 93.6 µm µm
Focal length = 100 mm vert.
Laser wavelength = 632 nm slit
Horizontal lines at edges of eyes remain.
Vertical lines above nose gone.
=
12 Robert R. McLeod, University of Colorado Fourier Optics 55 P3330 Exp Optics FA’2016
ECE 4606 Undergraduate Optics Lab
–Example
Simpler
Simpler objects
object
Low-pass
Original High-pass
A side note: this is how bandpass filters Low-pass, different
cutoff in x&y
look like in the frequency (focus) domain.
13 Robert R. McLeod, University of Colorado Fourier Optics 56 P3330 Exp Optics FA’2016
•Lecture 4
Phase contrast imaging
–Phase contrast
ECE 4606 Undergraduate Optics Lab
Phase contrast
REAL SPACE FOURIER SPACE
Ansel
− jπ
max ( Ansel )
e
Multiplied by
Filter cutoff frequency = 1/5000 µm-1
Filter cutoff position = 12.6 µm
Focal length = 100 mm
Laser wavelength = 632 nm Knife edge
Phase has become amplitude.
Zernike won the 1953 Nobel in Physics for this.
=
14 Robert R. McLeod, University of Colorado Fourier Optics 57 P3330 Exp Optics FA’2016
Some MATLAB tips: 2D functions
• Create a matrix that evaluates 2D
Gaussian: exp(-p/2(x2+y2)/s2)
– >>ind = [-32:1:31] / 32;
x – >>[x,y] = meshgrid(ind,-1*ind);
– >>z = exp(-pi/2*(x.^2+y.^2)/(.25.^2));
– >>imshow(z)
– >>colorbar
y y’
-1
64x64 matrix
x’
0
31/32
-1
15 Fourier Optics P3330 Exp Optics FA’2016
FFT “splits” low frequencies
1D Fourier Transform low frequencies (aka shift theorem)
12
– >>l = z(33,:); 10
abs (L)
– >>L = fft(l); 8
6
4
1
2
0.9
0
0.8 10 20 30 40 50 60
0.7
4
0.6
phase (L)
l
0.5 2
0.4
0
0.3
0.2 -2
0.1
-4
0 10 20 30 40 50 60
10 20 30 40 50 60
1 33 64 1 33 64
sample # sample #
0 xmax 0 fN/2 fN
x [Matlab]
16 Fourier Optics u P3330 Exp Optics FA’2016
Use FFTSHIFT prior to/after FFT or FFT2
Use fftshift for 2D functions
– >>smiley2 = fftshift(smiley);
17 Fourier Optics P3330 Exp Optics FA’2016
If your FFT looks jagged…
f = zeros(30,30);
f(5:24,13:17) = 1;
imshow(f,'InitialMagnification','fit')
F = fft2(f);
F2 = log(abs(F));
imshow(F2,[-1 5],'InitialMagnification','fit');
colormap(jet); colorbar
Discrete Fourier Transform Computed Without Padding
18 Fourier Optics P3330 Exp Optics FA’2016
Apply zero-padding!
F = fft2(f,256,256);
imshow(log(abs(F)),[-1 5]); colormap(jet);
colorbar
Discrete Fourier Transform With Padding
% Apply FFTSHIFT
F = fft2(f,256,256);F2 = fftshift(F);
imshow(log(abs(F2)),[-1 5]); colormap(jet);
colorbar
“wrong” low
Normal FT look
frequency location!
19 Fourier Optics P3330 Exp Optics FA’2016
Links/references
http://ecee.colorado.edu/~mcleod/teaching/ugol/lecturenotes/Lecture%2
04%20Fourier%20Optics.pdf
http://www.medphysics.wisc.edu/~block/bme530lectures/matlabintro.ppt
http://www.mathworks.com/help/images/fourier-transform.html
20 Fourier Optics P3330 Exp Optics FA’2016
