Reconstructing Multidimensional Signals from Projections | ECE 6258, Study notes of Digital Signal Processing

Material Type: Notes; Class: Digital Image Processing; Subject: Electrical & Computer Engr; University: Georgia Institute of Technology-Main Campus; Term: Fall 2003;

Typology: Study notes

Pre 2010

Uploaded on 08/05/2009

koofers-user-04o
koofers-user-04o 🇺🇸

5

(1)

9 documents

1 / 5

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
1
12/3/2003 ECE 6258 Russell M. Mersereau 1
ECE6258 Lecture 41
Reconstructing Multidimensional
Signals from Projections
12/3/2003 ECE 6258 Russell M. Mersereau 2
Motivation
A projection is like an “x-ray photograph” of a
multidimensional function.
The reconstruction of a signal from its projections is
equivalent to estimating a 3-D structure from a
series of x-ray photographs take from different
directions.
It is the only signal processing problem to be
awarded a Nobel Prize (Medicine, 1971)
12/3/2003 ECE 6258 Russell M. Mersereau 3
A Computed Tomography System
u
ˆ
object
u
ˆ
)
ˆ
(
1up
)
ˆ
(
2up
u
ˆ
)
ˆ
(
3up
12/3/2003 ECE 6258 Russell M. Mersereau 4
Formal Definition
Formally, a projection is a mapping from RNto RMfor
M<N.
In the special case
θ
=0
++= vdvuvuxup
vux
ˆ
)cos
ˆ
sin
ˆ
,sin
ˆ
cos
ˆ
()
ˆ
(
),(
444444344444421
θθθθ
θ
=vdvuxup ˆ
)
ˆ
,
ˆ
()
ˆ
(
0
pf3
pf4
pf5

Partial preview of the text

Download Reconstructing Multidimensional Signals from Projections | ECE 6258 and more Study notes Digital Signal Processing in PDF only on Docsity!

ECE 6258 Russell M. Mersereau

ECE6258 Lecture 41

Reconstructing MultidimensionalSignals from Projections

ECE 6258 Russell M. Mersereau

Motivation „

A

projection

is like an “x-ray photograph” of a

multidimensional function.

The reconstruction of a signal from its projections isequivalent to estimating a 3-D structure from aseries of x-ray photographs take from differentdirections.

It is the only signal processing problem to beawarded a Nobel Prize (Medicine, 1971)

ECE 6258 Russell M. Mersereau

A Computed Tomography System

u

object

u

) ˆ

( 1

u

p

) ˆ

( 2

u

p

u

) ˆ

( 3

u

p

ECE 6258 Russell M. Mersereau

Formal Definition „

Formally, a projection is a mapping from

R

N

to

R

M

for

M

N

In the special case

v d v u v u x u p

v

u

x

cos

sin

sin

cos

v d v u x u p

0

ECE 6258 Russell M. Mersereau

Applications „

X-ray photos (computed tomography)

„

Positron emission tomography (PET)

„

Electron microscopy

„

Marginal probability distributions

„

Fan-beam radio telescopes

„

Cross-Borehole measurements in seismicsignal processing

„

NMR imaging

ECE 6258 Russell M. Mersereau

The Problem

Given

for several

I

= 1,2,…,

P

, estimate

x

(

u

,

v

).

Approaches: ‰

Multidimensional Fourier transform

Radon Transform (spatial domain)

)

ˆ

(

u

p

i

ECE 6258 Russell M. Mersereau

Slices and Projections

ECE 6258 Russell M. Mersereau

Continuation

Therefore,

)

0 ,

(

)

(

μ

X

u

p

D

μ

ν

u

)

(

0

u

p

ECE 6258 Russell M. Mersereau

Playing with the sampling intervals „

By varying the sampling interval with the projectionangle, we can place samples on the same horizontaland vertical lines as the 2-D DFT.

ν

μ

ECE 6258 Russell M. Mersereau

Results (1-D Interpolation)

16 projections

32 projections 128 projections

64 projections

ECE 6258 Russell M. Mersereau

Radon Transform „

We begin with the inverse Fourier transform and convert the frequency

variables to polar coordinates.

Recall that

ECE 6258 Russell M. Mersereau

Radon Transform (2) „

This gives

ECE 6258 Russell M. Mersereau

Convolution Backprojection Algorithm „

This suggests the following algorithm

Filter each projection by |

ω

|.

Back project at the angle of the projection.

Add all the backprojections together.

Correct the DC value.

Iterative approach

ECE 6258 Russell M. Mersereau

Results

ECE 6258 Russell M. Mersereau

Results