Murari Image Processing Introduction - ENEA

Introduction to Image Processing
by Andrea Murari1, J.Vega2, T.Craciunescu3D.Mazon4,
L.Gabellieri5, M.Gelfusa6, D.Pacella5, A.Romano5
2
1
3
4
5
6 University of Rome
“Tor Vergata”
Images in History
Cameras and Images in Society
o One of the characteristics of modern societies is
their ability to easily capture and produce images
o Bidimensional detectors are very common: cellular
phones and CCTV cameras most visible applications
o The use of cameras and imaging has become much
more widespread in the entire fusion community
oThe new cameras for the operation of JET with the
ITER-like wall are more than 30.
oJET Fast Visible camera can produce Gbytes of data
per shot (comparable to a 2 hour digital movie)
oThe future of diagnostics for fusion is in the fields
of 2-D and even 3-D imaging
Various views on JET
Visible
IR
Plasma Boundary with SXR: the GEM Detector
•Pick-up coils have problems in a radiation hard environment
(close to the plasma, integrators etc)
•The next generation of plasmas will be so hot that even the
boundary will emit in the SXR
• Adapt Gas Electron Multiplier detectors
Cathode
Current~few nA
Anode
Preliminary tests on polycapillaries as SXR lenses
Full lens
Diverging beam
(magnification)
Half lens
Parallel beam
Full lens
X-Ray
source
5.65
cm
F
1
SXR radiography
of the mesh (holes 400 mm)
Medip
ix
4.4
cm
11.3
cm
F
2
27.5
cm
80 cm
SXR image of the mesh
with full
lens (magnification ~6 )
Humans and Image Processing
Human beings are well equipped to interpret visual data
(creature started using visual sensors in the Cambrian Age).
100 Million sensors in each human retina. Optical illusions
reveal the active role of the nervous system.
Main Issues
In the case of images captured by cameras what is
available is a bidimensional map of pixels with their
intensities.
Main challenges
a) Share amount of data
b) Ambiguity of visual
information
c) Extraction of
quantitative information (ill
posed problems)
See “Shape” by G.Stiny MIT Press.
Ill-posed Problems
If D is the space of the data or measurements, S
the source and A the forward function mapping the
reality on the space of the measurements
D= A(S)
The problem of recovering S from D is well posed
(accordig to the definition of Hadamard)
o A solution exists for any data D in data space
o The solution is unique in source (image) space S
o The inverse mapping D S is continuous
In general the vast majority on inverse problems
involving images in fusion are ill posed
Challenges of image processing
General topics:
o Information retrieval and confidence intervals
o Find solutions to reduce the ill-posed character of
the problems
o Extract efficiently physical information
oDevelop solutions compatible with real time control
Information processing capacity of the human brain:
o Conscious reasoning: 16 bits per second
o Mathematical reasoning: 12 bits per second
o Image processing : 10 Megabits per second