Title: Methods in Image Analysis
1Methods in Image Analysis Lecture 3Fourier
George Stetten, M.D., Ph.D.
- CMU Robotics Institute 16-725
- U. Pitt Bioengineering 2630
- Spring Term, 2004
2Frequency in time vs. space
- Classical signals and systems usually temporal
signals. - Image processing uses spatial frequency.
- We will review the classic temporal description
first, and then move to 2D and 3D space.
3Phase vs. Frequency
- Phase, , is angle, usually represented in
radians. - (circumference
of unit circle) - Frequency, , is the rate of change for phase.
- In a discrete system, the sampling frequency,
, is the amount of phase-change per sample.
4Eulers Identity
5Phasor Complex Number
6multiplication rotate and scale
7Spinning phasor
8(No Transcript)
9(No Transcript)
10(No Transcript)
11Continuous Fourier Series
is the Fundamental Frequency
Synthesis
Analysis
12Selected properties of Fourier Series
for real
13Differentiation boosts high frequencies
14Integration boosts low frequencies
15Continuous Fourier Transform
Synthesis
Analysis
16Selected properties of Fourier Transform
17Special Transform Pairs
- Impulse has all frequences
- Average value is at frequency 0
- Aperture produces sync function
18Discrete signals introduce aliasing
Frequency is no longer the rate of phase change
in time, but rather the amount of phase change
per sample.
19Sampling gt 2 samples per cycle
20Sampling lt 2 samples per cycle
21Under-sampled sine
22Discrete Time Fourier Series
Sampling frequency is 1 cycle per second, and
fundamental frequency is some multiple of that.
Synthesis
Analysis
23Matrix representation
24Fast Fourier Transform
- N must be a power of 2
- Makes use of the tremendous symmetry within the
F-1 matrix - O(N log N) rather than O(N2)
25Discrete Time Fourier Transform
Sampling frequency is still 1 cycle per second,
but now any frequency are allowed because xn is
not periodic.
Synthesis
Analysis
26The Periodic Spectrum
27Aliasing Outside the Base Band
Perceived as
282D Fourier Transform
Analysis
or separating dimensions,
Synthesis
29Properties
- Most of the usual properties, such as linearity,
etc. - Shift-invariant, rather than Time-invariant
- Parsevals relation becomes Rayleighs Theorem
- Also, Separability, Rotational Invariance, and
Projection (see below)
30Separability
31Rotation Invariance
32Projection
Combine with rotation, have arbitrary projection.
33Gaussian
seperable
Since the Fourier Transform is also separable,
the spectra of the 1D Gaussians are, themselves,
separable.
34Hankel Transform
For radially symmetrical functions
35Variable Conductance Diffusion (VCD)
- Attempt to get around the global nature of
Fourier. - Smoothing with a Gaussian in the spatial domain
yields multiplication by a Gaussian in the
frequency domain, i.e., a low pass filter. - This lowers noise, but also blurs boundaries.
- Gaussian smoothing simulates uniform heat
diffusion. - VCD makes conductance an inverse function of
gradient, so that heat does not flow well
across boundaries. This homogenizes already
homogenious regions while preserving boundaries.
36Elliptical Fourier Series for 2D Shape
Parametric function, usually with constant
velocity.
Truncate harmonics to smooth.
37Fourier shape in 3D
- Fourier surface of 3D shapes (parameterized on
surface). - Spherical Harmonics (parameterized in spherical
coordinates). - Both require coordinate system relative to the
object. How to choose? Moments? - Problem of poles singularities cannot be avoided
38Quaternions 3D phasors
Product is defined such that rotation by
arbitrary angles from arbitrary starting points
become simple multiplication.
39Summary
- Fourier useful for image processing,
convolution becomes multiplication. - Fourier less useful for shape.
- Fourier is global, while shape is local.
- Fourier requires object-specific coordinate
system.