fourier analysis without tears
DESCRIPTION
Fourier Analysis Without Tears. Somewhere in Cinque Terre, May 2005. 15-463: Computational Photography Alexei Efros, CMU, Fall 2006. x 1. x 1. 1 st principal component. 2 nd principal component. x 0. x 0. Capturing what’s important. Fast vs. slow changes. A nice set of basis. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/1.jpg)
Fourier Analysis Without Tears
15-463: Computational PhotographyAlexei Efros, CMU, Fall 2006
Somewhere in Cinque Terre, May 2005
![Page 2: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/2.jpg)
Capturing what’s important
x1
x0
x1
x0
1st principal component2nd principal
component
![Page 3: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/3.jpg)
Fast vs. slow changes
![Page 4: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/4.jpg)
A nice set of basis
This change of basis has a special name…
Teases away fast vs. slow changes in the image.
![Page 5: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/5.jpg)
Jean Baptiste Joseph Fourier (1768-1830)had crazy idea (1807):
Any periodic function can be rewritten as a weighted sum of sines and cosines of different frequencies.
Don’t believe it? • Neither did Lagrange,
Laplace, Poisson and other big wigs
• Not translated into English until 1878!
But it’s true!• called Fourier Series
![Page 6: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/6.jpg)
A sum of sinesOur building block:
Add enough of them to get any signal f(x) you want!
How many degrees of freedom?
What does each control?
Which one encodes the coarse vs. fine structure of the signal?
xAsin(
![Page 7: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/7.jpg)
Fourier TransformWe want to understand the frequency of our signal. So, let’s reparametrize the signal by instead of x:
xAsin(
f(x) F()Fourier Transform
F() f(x)Inverse Fourier Transform
For every from 0 to inf, F() holds the amplitude A and phase of the corresponding sine
• How can F hold both? Complex number trick!
)()()( iIRF 22 )()( IRA
)()(tan 1
RI
We can always go back:
![Page 8: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/8.jpg)
Time and Frequencyexample : g(t) = sin(2pf t) + (1/3)sin(2p(3f) t)
![Page 9: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/9.jpg)
Time and Frequencyexample : g(t) = sin(2pf t) + (1/3)sin(2p(3f) t)
= +
![Page 10: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/10.jpg)
Frequency Spectraexample : g(t) = sin(2pf t) + (1/3)sin(2p(3f) t)
= +
![Page 11: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/11.jpg)
Frequency SpectraUsually, frequency is more interesting than the phase
![Page 12: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/12.jpg)
= +
=
Frequency Spectra
![Page 13: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/13.jpg)
= +
=
Frequency Spectra
![Page 14: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/14.jpg)
= +
=
Frequency Spectra
![Page 15: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/15.jpg)
= +
=
Frequency Spectra
![Page 16: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/16.jpg)
= +
=
Frequency Spectra
![Page 17: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/17.jpg)
= 1
1 sin(2 )k
A ktk
Frequency Spectra
![Page 18: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/18.jpg)
Frequency Spectra
![Page 19: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/19.jpg)
FT: Just a change of basis
.
.
.
* =
M * f(x) = F()
![Page 20: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/20.jpg)
IFT: Just a change of basis
.
.
.
* =
M-1 * F() = f(x)
![Page 21: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/21.jpg)
Finally: Scary Math
![Page 22: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/22.jpg)
Finally: Scary Math
…not really scary:is hiding our old friend:
So it’s just our signal f(x) times sine at frequency
)sin()cos( xixe xi
QPQPΑ
xAxQxP
122 tan
sin()sin()cos(
xAsin(phase can be encoded
by sin/cos pair
![Page 23: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/23.jpg)
Extension to 2D
in Matlab, check out: imagesc(log(abs(fftshift(fft2(im)))));
![Page 24: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/24.jpg)
2D FFT transform
![Page 25: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/25.jpg)
Man-made Scene
![Page 26: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/26.jpg)
Can change spectrum, then reconstruct
![Page 27: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/27.jpg)
Most information in at low frequencies!
![Page 28: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/28.jpg)
Campbell-Robson contrast sensitivity curveCampbell-Robson contrast sensitivity curve
We don’t resolve high frequencies too well…
… let’s use this to compress images… JPEG!
![Page 29: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/29.jpg)
Lossy Image Compression (JPEG)
Block-based Discrete Cosine Transform (DCT)
![Page 30: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/30.jpg)
Using DCT in JPEG A variant of discrete Fourier transform
• Real numbers• Fast implementation
Block size• small block
– faster – correlation exists between neighboring pixels
• large block– better compression in smooth regions
![Page 31: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/31.jpg)
Using DCT in JPEG The first coefficient B(0,0) is the DC component,
the average intensityThe top-left coeffs represent low frequencies,
the bottom right – high frequencies
![Page 32: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/32.jpg)
Image compression using DCTDCT enables image compression by
concentrating most image information in the low frequencies
Loose unimportant image info (high frequencies) by cutting B(u,v) at bottom right
The decoder computes the inverse DCT – IDCT
•Quantization Table3 5 7 9 11 13 15 175 7 9 11 13 15 17 197 9 11 13 15 17 19 219 11 13 15 17 19 21 2311 13 15 17 19 21 23 2513 15 17 19 21 23 25 2715 17 19 21 23 25 27 2917 19 21 23 25 27 29 31
![Page 33: Fourier Analysis Without Tears](https://reader030.vdocuments.us/reader030/viewer/2022033012/5681609d550346895dcfc569/html5/thumbnails/33.jpg)
JPEG compression comparison
89k 12k