Professional Documents
Culture Documents
Fourier Transform II
Fourier Transform II
Fourier Transform II
Fourier Transform Ⅱ
Signal in an image:
Optical density on radiograph
Brightness on display monitor
f (x)
Signal
f (x)
Signal
x (mm) x (mm)
0 x (cm)
Fourier Transform
∞ 1 ∞ 1 ∞
∫ ∫ ∫
−iωξ iωx
f ( x) = [ f (ξ )e dξ ]e dω = F (ω )e iωx dω
− ∞ 2π −∞ 2π −∞
∞
F (ω ) = ∫ −∞
f ( x)e −iωx dx Fourier Transform (2.24)
1 ∞
f ( x) =
2π ∫
−∞
F (ω )e iωx dω Inverse Fourier Transform (2.25)
f(x) F(u)
FT
Inverse FT
x (cm) u (cycles/cm)
FT
Inverse FT
∞ 1 ∞
∫ f ( x) dx = ∫ F (ω ) dω ; Parseval' s Theorem
2 2
−∞ 2π −∞
F (ω )
2
; Power Spectrum
f(x)
1/2d ⎧1 / 2d ( x ≤ d)
f ( x) = ⎨
⎩0 ( x > d)
-d d x
∞
From (2.24), F (ω ) = ∫
−∞
f ( x)e −iωx dx
1 −iωx
d 1 1 −iωx d
= ∫
− d 2d
e dx = −
2 d iω
[e ]− d
1 e i ωd − e − i ωd
=
ωd 2i
sin( ωd ) e iθ − e −iθ
= (sin(θ ) = )
ωd 2i
= sin c(ωd )
Fourier Transform of Rectangle Pulse 2
f(x) F(ω)
1
1/2d
FT
-d d ω
x
−2π/d −π/d
π/d 2π/d
⎧1 / 2d ( x ≤ d) sin( ωd )
f ( x) = ⎨ F (ω ) = = sin c(ωd )
⎩0 ( x > d) ωd
f(x) F(ω)
1
FT
ω
x
wide width narrow width
(low freq. component)
F(ω)
f(x)
1
FT
ω
x
narrow width wide width
(high freq. component)
Fourier Transform of Sinc Function
F(x) f(ω)
1
1/2d
FT
x -d d ω
−2π/d −π/d
π/d 2π/d
sin( xd ) ⎧1 / 2d (ω ≤ d)
F ( x) = = sin c( xd ) f (ω ) = ⎨
xd ⎩0 (ω > d)
0 a
x ∫
−∞
f ( x)δ ( x)dx = f (0)
∞
∫
−∞
f ( x)δ ( x − a )dx = f (a )
∞
∫
−iωx − iω 0
F (ω ) = δ ( x )e dx = e = e =1
0
−∞
Fourier Transform of δ-Function 2
F(ω )
δ(x)
FT
1
x 0 ω
0
δ(x-x0)
∞
ℑ[δ ( x − x0 )] = ∫
−∞
δ ( x − x0 )e −iωx dx = e −iωx0
0 x0 x
Fourier Transform of δ-Function Sequence 1
f(x)
δ-function sequence
(or comb function):
f ( x) = L + δ ( x + 2 x0 ) + δ ( x + x0 ) + δ ( x) + δ ( x − x0 ) + +δ ( x − 2 x0 )L
∞
= ∑ δ ( x − nx )
n = −∞
0
Fourier Transform of δ-Function Sequence 2
∞ ∞
f ( x) = ∑
n = −∞
cn e inω0 x
= ∑
n = −∞
cn e i 2πnu0 x (ω0 = 2πu0 )
where, u0 = 1 / x0
1 x0 / 2
cn =
x0 ∫− x0 / 2
f ( x)e −i 2πnu0 x dx
1 x0 / 2 1 0
=
x0 ∫ − x0 / 2
δ ( x)e −i 2πnu0 x dx =
x0
e
1
= = u0
x0
Therefore,
∞ ∞
f ( x) = ∑u e
n = −∞
0
i 2πnu0 x
= u0 ∑e
n = −∞
i 2πnu0 x
Fourier Transform of δ-Function Sequence 3
∞
F (u ) = ℑ[u0 ∑e
n = −∞
i 2πnu0 x
]
∞
= u0 ∑ ℑ[e
n = −∞
i 2πnu0 x
]
Therefore,
ℑ[e i 2πnu0 x ] = δ (−u + nu0 ) = δ (u − nu0 ) (even function: δ (U ) = δ (−U ))
Finally, ∞
F (u ) = u0 ∑ δ (u − nu )
n = −∞
0
Fourier Transform of δ-Function Sequence 4
∞ ∞
ℑ[ ∑ δ ( x − nx )] = u ∑ δ (u − nu ),
n = −∞
0 0
n = −∞
0 where u0 = 1 / x0
f(x) F(u)
FT
x u
0 0
x0 u0 =1/x0
∞ ∞
f ( x) = ∑ δ ( x − nx )
n = −∞
0 F (u ) = u0 ∑ δ (u − nu )
n = −∞
0
Screen
Imaging Plate Flat Panel Detector
Film
Digitization Digitization
Film Processing
system system system
Image Image Image
Screen/Film CR DR
Input-Output System for Image Formation
input output
System
transmitted x-ray image
f(x) g(x)
g ( x ) = L[ f ( x )]
where L indicate a conversion rule from f(x) to g(x).
input output
System
f(x) g(x)
Additivity
Steadiness
L[ f ( x − x1 )] = g ( x − x1 )
optical density
because film has not additivity.
exposure
Linear System Response 1
f(x) g(x)
Linear System
∞ ∞
f ( x) = ∫ f (τ )δ ( x − τ )dτ because ∫ f ( x)δ ( x)dx = f (0)
−∞ −∞
δ(x) h(x)
h( x) = L[δ ( x)]
x x
Linear System
Impulse Response
Linear System Response 2
f(x) g(x)
Linear System
g ( x) = L[ f ( x)]
∞
= L[ ∫ f (τ )δ ( x − τ )dτ ]
−∞
∞
= ∫ f (τ ) L[δ ( x − τ )]dτ
−∞
∞
= ∫ f (τ )h( x − τ )dτ
−∞
= f ( x ) * h( x )
Linear System Response 3
g ( x ) = f ( x ) * h( x )
if F (ω ) = ℑ[ f ( x)], G (ω ) = ℑ[ g ( x)], and H (ω ) = ℑ[h( x)],
from convolution integral theorem,
G (ω ) = F (ω ) H (ω )
h(x)
f (x ) g ( x) = f ( x) * h( x ) :in real domain
Linear System
F (ω ) G (ω ) = F (ω ) H (ω ) :in frequency domain
H (ω )
Linear System Response in Real Domain
h(x)
f (x ) g ( x) = f ( x ) * h( x)
Linear System
input:f(x) output:g(x)
f(τ)δ(x-τ) f(τ)h(x-τ)
x x
δ(x) h(x)
x x
Linear System h(x): impulse response
δ(x) h(x)
ℑ[δ ( x)]
H (u ) MTF (modulation
transfer function)
u
u
Modulation Transfer Function (MTF) 1
δ (x) h(x)
Linear System
H (u ) = ℑ[ h( x )]
ℑ[δ ( x)]
1.0
1.0
0.8
u | H (u ) |
MTF
0.4
0.2
0.0
0 1 2 3 4 u(cycles/mm)
Modulation Transfer Function (MTF) 2
A B
System System
1 mm
1 mm
1 mm u=1 cycles/mm 1 mm u=2 cycles/mm
C 1.0
A
0.8
| H (u ) |
System
MTF
B
0.4
1.0 0.2
C
1 mm 0.2
1 mm u=4 cycles/mm 0.0
0 1 2 3 4 u(cycles/mm)
Two Dimensional (2D) Fourier Transform
∞ ∞
1D FT F (u ) = ∫ f ( x)e −i 2πux dx, f ( x) = ∫ F (u )ei 2πux du
−∞ −∞
F (u , v) = ℑ[ f ( x, y )]
∞ ∞
F (u , v) = ∫ ∫ f ( x, y )e −i 2π ( ux+ vy ) dxdy
−∞ −∞
∞ ∞
f ( x, y ) = ∫ ∫ F (u , v)e i 2π (ux+ vy ) dudv
−∞ −∞
Two Dimensional (2D) Discrete Fourier Transform (DFT)
origin of an image
u
4 3
v
2 1
u
3 4
1 2