5.11 m Image Reconstruction from Projections 251
PSF estimation is affected strongly by the size of its initial guess, and
SS values (an array of Is is a reasonable starting guess),
amber of iterations performed with the preceding syntax is 10 by
Additional parameters may be included in the function to control the
iterations and other features of the restoration, as in the following
| PSF] = deconvblind(g, INITPSF, NUMIT, DANPAR, WEIGHT)
MUUIT, DAMPAR, and WETGHT are as described for the L-R algorithm in
sous section,
|= restored image exhibits ringing introduced by the discrete Fourier
= used in the algorithm, it helps sometimes to use function edgetaper’
of L-D projections This problem, typically called computed tomography
is one of the principal applications of image processing in medicine,
EXAMPLES
Using finetion
Gecenvelins to
estimate a PSF7
252. Ghapler 5m Image Restoration and Reconstruction
FIGURE 5.11
{a) Original PSF.
(b) through (d)
Estimates of the
PSF using 5,10,
‘nd 20 iterations
in function
‘deconvb ind.
s no way of deterr
single object or
onstruction bi
mage by proiec
iginal beam, a8 F
cates a 2-D digital i
mage is of little v
5.11.) Background
The foundation of image reconstruction from projections is straight
aoe venibe explained intuitively. Consider the region in Fig, 512(a).T
physica meaning tothe following discussion, assume that this regions
Fough a section of a human body showing a tumor (bright area)
oye mnogencous area of tissue (black background). Such a region mis
in tained for example, by passing a thin, fat beam of X-rays perpencl
cre touy, and recording at the opposite end measurements proporto srangement by %
the absorption of the beam as it passes through the body. The tumor abs adding the resul
see ar the X-ray enerey,hence giving a higher reading for absorption. = = Fig 5.12(¢)-Note ho
igual (absorption profile) onthe riht side of Fg. 5.12(a) shows Obser
ae ham abvorption occurs through the center ofthe region, where the Bea
aaa ners the longest path through the tumor. At this point, the abso
profi is all the information we have about the object,
intensity of the 0
sitive that we shou
~ backprojections
what happens. As5.11 a Image Reconstruction from Projections 253
Absorption pratile —
Ray Detectorstip —#
ea tet
ETT LUNI
way of determining from a single projection whether we are
Single object or multiple objects along the path of the beam, but
onstruction based on this partial information. The approach is
image by projecting the absorption profile back along the direc.
iginal beam, as Fig. 5.12(b) shows This process, called backprojec.
5 82:D digital image from a I-D absorption profile waveform. By
is of litle value. However, suppose that we rotate the beam!
angement by 90° [Fig. 5.12(c)] and repeat the backprojection
dding the resulting backprojection to Fig. 5.12(b) we obtain the
=-5.12(e). Note how the intensity of the region containing the abject
tensity of the other major components of the image,
ve that we should be able to refine the preceding results by gener.
ackprojections at different angles, As Figs, 5.12(f)-(h) show. this is
at happens. As the number of backprojections increases, the area
ab
ede
teh
une 5.12
(@) Fatregion
vith object,
Paallel beam
{etector stip and
absorption profile
{b) Backprojcton
ot absorption
prof. c) Beam
and detector st
rotated 90° and .
(6) Backprojetion
of absorption
profile. @) Sum
1 (6) and (2).
Result of adding
another
backprojection (at
455) @) Rewltor
acing yet another
backprojection at
1357 (h) Result of
backprojeetions
5
apart,