How to decompose a signal into sine and cosine function Also known as harmonic functions ▫ Fourier Transform, Discrete Fourier Transform, Discrete Cosine
Previous PDF | Next PDF |
[PDF] Table of Fourier Transform Pairs
Function, f(t) Fourier Transform, F(w) Definition of Inverse Fourier Transform Р cos( t t p t rect t A 2 2 )2( ) cos( w t p wt t p - A ) cos( 0t w [ ]) () ( 0 0 wwd
[PDF] Fourier Transform - Rutgers CS - Rutgers University
How to decompose a signal into sine and cosine function Also known as harmonic functions ▫ Fourier Transform, Discrete Fourier Transform, Discrete Cosine
[PDF] Discrete Fourier Transform Inverse Fourier transform
12 avr 2018 · The point of the Fourier transform is to be able to write a function as a sum of sinuosoids Since sine and cosine functions are defined over all
[PDF] Fourier Transforms of Common Functions - CS-UNM
The inverse Fourier transform transforms a func- tion of frequency, F(s), into a function of time, f(t): F −1 Fourier Transform of Sine and Cosine (contd )
[PDF] The Fourier Transform: Examples, Properties, Common Pairs
Example: Fourier Transform of a Cosine f(t) = cos(2πst) Odd and Even Functions Even Odd Let F−1 denote the Inverse Fourier Transform: f = F−1(F )
[PDF] Addl Table of Fourier Transform Pairs/Properties
Signals Systems - Reference Tables 1 Table of Fourier Transform Pairs Function, f(t) Fourier Transform, F(w) Definition of Inverse Fourier Transform ò ¥ ¥-
[PDF] Chapter 3 - Sine and Cosine Transforms - WordPresscom
Transforms with cosine and sine functions as the transform kernels represent an used here provides for a definition for the inverse Fourier cosine transform,
[PDF] Chapter 1 The Fourier Transform - Math User Home Pages
1 mar 2010 · cos(λt)dt = 2 sin(πλ) λ = 2π sinc λ Thus sinc λ is the Fourier transform of the box function The inverse Fourier transform is ∫ ∞ −
[PDF] 13 Fourier Transform
An ”integral transform” is a transformation that produces from given functions new func- tions that Inverse Fourier cosine transform of ˆfc(ω): f(x) = √ 2 π ∫ ∞
[PDF] Chapter 4 Fourier Transform
is di erentiable everywhere, however, g0(x)=2xsin(1/x)¡cos(1/x) and thus g0(0+) The Fourier transform is usually de ned for admissible functions, and for this The following de nition gives an inverse relation to the Fourier transform
[PDF] inverse fourier transform of rectangular function
[PDF] inverse fourier transform of step function
[PDF] inverse z transform calculator
[PDF] inverse z transform of 1
[PDF] investigation 2 i check mixtures and solutions answers
[PDF] investment protection agreement
[PDF] investors expedia group
[PDF] inwi *5
[PDF] inwi data recharge
[PDF] inwi promotion
[PDF] inwi recharge *6
[PDF] ioc host city contract
[PDF] iodoform test
[PDF] ios app development project plan
1
CS443: Digital Imaging and Multimedia
Introduction to Spectral Techniques
Spring 2008
Ahmed Elgammal
Dept. of Computer Science
Rutgers University
Outlines
yFourier Series and Fourier integral yFourier Transform (FT) yDiscrete Fourier Transform (DFT) yAliasing and Nyquest Theorem y2D FT and 2D DFT yApplication of 2D-DFT in imaging yInverse Convolution yDiscrete Cosine Transform (DCT)Sources:
yBurger and Burge "Digital Image Processing" Chapter 13, 14, 15 yFourier transform images from Prof. John M. Brayer @ UNM 2 yRepresentation and Analysis of Signals in the frequency domain yAudio: 1D temporal signal yImages: 2D spatial signal yVideo: 2D spatial signal + 1D temporal signal yHow to decompose a signal into sine and cosine function. Also known as harmonic functions. yFourier Transform, Discrete Fourier Transform,Discrete Cosine Transform
Basics
ySine and Cosine functions are periodic yAngular Frequency: number of oscillations over the distance 2πT: the time for a complete cycle
3Basics
yAngular Frequency (ω) and Amplitude (a) yAngular Frequency: number of oscillations over the distance 2πT: the time for a complete cycle
yCommon Frequency f: number of oscillation in a unit timeBasics
yPhase: Shifting a cosine function along the x axis by a distance ϕ change the phase of the cosine wave. ϕ denotes the phase angle 4 yAdding cosines and sines with the same frequency results in another sinusoidFourier Series and Fourier integral
yWe can represent any periodic function as sum of pairs of sinusoidal functions- using a basic (fundamental) frequency yFourier Integral: any function can be represented as combination of sinusoidal functions with many frequencies 5 yFourier Integral yHow much of each frequency contributes to a given functionFourier Transform
6 yFourier transform yInverse Fourier transformFrequency domain
FTTemporal or spatial domain
Fourier Transform
yThe forward and inverse transformation are almost similar (only the sign in the exponent is different) yany signal is represented in the frequency space by its frequency "spectrum" yThe Fourier spectrum is uniquely defined for a given function. The opposite is also true. yFourier transform pairs 7 8 9 ySince of FT of a real function is generally complex, we use magnitude and phaseLower frequencies ⇒ narrower power spectrum
Higher frequencies ⇒ wider power spectrum
u |F(u)| 2 x f(x)Power Spectrum
10Properties
ySymmetry: for real-valued functions yLinearity ySimilarity yShift Property 11Important Properties:
yFT and Convolution yConvolving two signals is equivalent to multiplying theirFourier spectra
y Multiplying two signals is equivalent to convolving theirFourier spectra
yFT of a Gaussian is a GaussianDiscrete Fourier Transform
yIf we discretize f(x) using uniformly spaced samples f(0), f(1),...,f(N-1), we can obtain FT of the sampled function yImportant Property:Periodicity F(m)=F(m+N)
One period
12 Image from Computer Graphics: Principles and Practice by Foley, van Dam, Feiner, and HughesImpulse function
13 14Sampling and Aliasing
yDifferences between continuous and discrete images yImages are sampled version of a continuous brightness function. successful sampling unsuccessful sampling 15Sampling and Aliasing
ySampling involves loss of information yAliasing: high spatial frequency components appear as low spatial frequency components in the sampled signal successful sampling unsuccessful sampling Java applet from: http://www.dsptutor.freeuk.com/aliasing/AD102.htmlAliasing
yNyquist theorem: The sampling frequency must be at least twice the highest frequency present for a signal to be reconstructed from a sampled version. (Nyquist frequency) 16Temporal Aliasing
A AA A AThe wheel appears to be moving backwards at about
angular frequency 17Sampling, aliasing, and DFT
yDFT consists of a sum of copies of the FT of the original signal shifted by the sampling frequency: yIf shifted copies do not intersect: reconstruction is possible. yIf shifted copies do intersect: incorrect reconstruction, high frequencies are lost (Aliasing)2-dimension
In two dimension
•These terms are sinusoids on the x,y plane whose orientation and frequency are defined by u,v 18DFT in 2D
yFor a 2D periodic function of size MxN, DFT is defined as: yInverse transform 19 20Visualizing 2D-DFT
yThe FT tries to represent all images as a summation of cosine-like imagesImages of pure cosines
FT •Center of the image:the origin of the frequency coordinate system •u-axis: (left to right) the horizontal component of frequency •v-axis: (bottom-top) the vertical component of frequency •Center dot (0,0) frequency : image average •high frequencies in the vertical direction will cause bright dots away from the center in the vertical direction. •high frequencies in the horizontal direction will cause bright dots away from the center in the horizontal direction. ySince images are real numbers (not complex) FT image is symmetric around the origin.FT: symmetry
FT is shift invariant
21yIn general, rotation of the image results in equivalent rotation of its FT
Why it is not the case ?
•Edge effect ! •FT always treats an image as if it were part of a periodically replicated array of identical images extending horizontally and vertically to infinity •Solution: "windowing" the imageEdge effect
22•notice a bright band going to high frequencies perpendicular to the strong edges in the image •Anytime an image has a strong- contrast, sharp edge the gray values must change very rapidly.
It takes lots of high frequency
power to follow such an edge so there is usually such a line in its magnitude spectrum. 23ScalingPeriodic image patterns
24RotationOriented, elongated structures
25Natural Images
•notice a bright band going to high frequencies perpendicular to the strong edges in the image •Anytime an image has a strong- contrast, sharp edge the gray values must change very rapidly.It takes lots of high frequency
power to follow such an edge so there is usually such a line in its magnitude spectrum. 26Print patternsLinear Filters in Frequency space
27Inverse Filters - De-convolution
yHow can we remove the effect of a filter ?Inverse Filters - De-convolution
yHow can we remove the effect of a filter ? 28yWhat happens if we swap the magnitude spectra ? yPhase spectrum holds the spatial information (where things are), yPhase spectrum is more important for perception than magnitude spectrum.