| 1 |
\chapter{Concepts} |
| 2 |
\setheader{{\it CHAPTER \thechapter}}{}{}{\ctsimheadtitle}{}{{\it CHAPTER \thechapter}} |
| 3 |
\ctsimfooter |
| 4 |
|
| 5 |
\section{Conceptual Overview}\index{Conceptual overview} |
| 6 |
The operation of \ctsim\ begins with the phantom object. A |
| 7 |
phantom object consists of geometric elements. A scanner is |
| 8 |
specified and the collection of x-ray data, or projections, is |
| 9 |
simulated. This projection data can be reconstructed using various |
| 10 |
user-controlled algorithms producing an image of the phantom |
| 11 |
object. These reconstructions can be visually and statistically |
| 12 |
compared to the original phantom object. |
| 13 |
|
| 14 |
In order to use \ctsim\ effectively, some knowledge of how |
| 15 |
\ctsim\ works and the approach taken is required. \ctsim\ deals with a |
| 16 |
variety of object, but the two primary objects that we need to be |
| 17 |
concerned with are the \helprefn{phantom}{conceptphantom} and the |
| 18 |
\helprefn{scanner}{conceptscanner}. |
| 19 |
|
| 20 |
\section{Phantoms}\label{conceptphantom} |
| 21 |
|
| 22 |
\ctsim\ uses geometrical objects to describe the object being |
| 23 |
scanned. A phantom is composed of one or more phantom elements. |
| 24 |
These elements are simple geometric shapes, specifically, |
| 25 |
rectangles, triangles, ellipses, sectors and segments. With these |
| 26 |
elements, the standard phantoms used in the CT literature can be |
| 27 |
constructed. In fact, \ctsim\ provides a shortcut to load the |
| 28 |
published phantoms of Herman\cite{HERMAN80} and |
| 29 |
Shepp-Logan\cite{SHEPP74}. \ctsim\ also reads text files of |
| 30 |
user-defined phantoms. |
| 31 |
|
| 32 |
The types of phantom elements and their definitions are taken with |
| 33 |
permission from G.T. Herman's publication\cite{HERMAN80}. |
| 34 |
|
| 35 |
\subsection{Phantom File}\label{phantomfile}\index{Phantom!File syntax} |
| 36 |
Each line in the text file describes an element of the |
| 37 |
phantom. Each line contains seven entries, in the following form: |
| 38 |
\begin{verbatim} |
| 39 |
element-type cx cy dx dy r a |
| 40 |
\end{verbatim} |
| 41 |
The first entry defines the type of the element, either |
| 42 |
\texttt{rectangle}, \texttt{ellipse}, \texttt{triangle}, |
| 43 |
\texttt{sector}, or \texttt{segment}. |
| 44 |
|
| 45 |
For all phantom elements, \texttt{r} is the rotation applied to the object in degrees |
| 46 |
counterclockwise and \texttt{a} is the X-ray attenuation |
| 47 |
coefficient of the object. Where objects overlap, the attenuations |
| 48 |
of the overlapped objects are summed. |
| 49 |
|
| 50 |
As opposed to the \texttt{r} and \texttt{a} fields, the \texttt{cx}, |
| 51 |
\texttt{cy}, \texttt{dx} and \texttt{dy} fields have different |
| 52 |
meanings depending on the element type. |
| 53 |
|
| 54 |
|
| 55 |
|
| 56 |
\subsection{Phantom Elements}\label{phantomelements}\index{Phantom!Elements} |
| 57 |
|
| 58 |
\subsubsection{ellipse} |
| 59 |
Ellipses use \texttt{dx} and \texttt{dy} to define the semi-major and |
| 60 |
semi-minor axis lengths with the center of the ellipse at \texttt{(cx,cy)}. |
| 61 |
Of note, the commonly used phantom described by |
| 62 |
Shepp and Logan\cite{SHEPP74} uses only ellipses. |
| 63 |
|
| 64 |
\subsubsection{rectangle} |
| 65 |
Rectangles use \texttt{(cx,cy)} to define the position of |
| 66 |
the center of the rectangle with respect to the origin. \texttt{dx} |
| 67 |
and \texttt{dy} are the half-width and half-height of the rectangle. |
| 68 |
|
| 69 |
\subsubsection{triangle} |
| 70 |
Triangles are drawn with the center of the base at \texttt{(cx,cy)} |
| 71 |
and a base half-width of \texttt{dx} and a height of \texttt{dy}. |
| 72 |
Rotations are then applied about the center of the base. |
| 73 |
|
| 74 |
\subsubsection{segment} |
| 75 |
Segments are complex. They are the portion of an circle between a |
| 76 |
chord and the perimeter of the circle. \texttt{dy} sets the |
| 77 |
radius of the circle. Segments start with the center of the chord |
| 78 |
located at \texttt{(0,0)} and the chord horizontal. The half-width |
| 79 |
of the chord is set by \texttt{dx}. The portion of an circle |
| 80 |
lying below the chord is then added. The imaginary center of this |
| 81 |
circle is located at \texttt{(0,-dy)}. The segment is then rotated |
| 82 |
by \texttt{r} and then translated by \texttt{(cx,cy)}. |
| 83 |
|
| 84 |
\subsubsection{sector} |
| 85 |
Sectors are the like a ``pie slice'' from a circle. The radius of |
| 86 |
the circle is set by \texttt{dy}. Sectors are defined similarly to |
| 87 |
segments. In this case, though, a chord is not drawn. Instead, |
| 88 |
the lines are drawn from the origin of the circle \texttt{(0,-dy)} |
| 89 |
to the points \texttt{(-dx,0)} and \texttt{(dx,0)}. The perimeter |
| 90 |
of the circle is then drawn between those two points and lies |
| 91 |
below the x-axis. The sector is then rotated and translated the |
| 92 |
same as a segment. |
| 93 |
|
| 94 |
\subsection{Phantom Size}\index{Phantom!Size} |
| 95 |
The overall dimensions of the phantom are increased by 1\% above the |
| 96 |
specified sizes to avoid clipping due to round-off errors from |
| 97 |
sampling the polygons of the phantom elements. So, if the phantom is |
| 98 |
defined as a rectangle of size 0.1 by 0.1, the phantom size is |
| 99 |
0.101 in each direction. |
| 100 |
|
| 101 |
\section{Scanner}\label{conceptscanner}\index{Scanner!Concepts}% |
| 102 |
Understanding the scanning geometry is the most complicated aspect of |
| 103 |
using \ctsim. For real-world CT simulators, this is actually quite |
| 104 |
simple. The geometry is fixed by the manufacturer during the |
| 105 |
construction of the scanner and can not be changed. \ctsim, |
| 106 |
being a very flexible simulator, gives tremendous options in |
| 107 |
setting up the geometry for a scan. |
| 108 |
|
| 109 |
\subsection{Dimensions} |
| 110 |
The geometry for a scan starts with the size of |
| 111 |
the phantom being scanned. This is because \ctsim\ allows for |
| 112 |
statistical comparisons between the original phantom image and |
| 113 |
it's reconstructions. Since CT scanners scan a circular area, the |
| 114 |
first important variable is the diameter of the circle surround |
| 115 |
the phantom, the \emph{phantom diameter}. Remember, as mentioned |
| 116 |
above, the phantom dimensions are padded by 1\%. |
| 117 |
|
| 118 |
The other important geometry variables for scanning phantoms are |
| 119 |
the \emph{view diameter}, \emph{scan diameter}, \emph{focal |
| 120 |
length}, and \emph{center-detector length}. These variables are input into \ctsim\ in terms of |
| 121 |
ratios rather than absolute values. |
| 122 |
|
| 123 |
\subsubsection{Phantom Diameter}\index{Phantom!Diameter} |
| 124 |
The phantom diameter is automatically calculated by \ctsim\ from |
| 125 |
the phantom definition. The maximum of the phantom length and |
| 126 |
height is used to define the square that completely surrounds the |
| 127 |
phantom. Let \latexonly{$p_l$}\latexignore{\emph{Pl}} be the width |
| 128 |
and height of this square. The diameter of this boundary box, |
| 129 |
\latexonly{$p_d$,}\latexignore{\emph{Pd},} is given by the |
| 130 |
Pythagorean theorem and is |
| 131 |
\latexignore{\\\centerline{\emph{Pl x sqrt(2)}}\\} |
| 132 |
\latexonly{\begin{equation}p_d = p_l \sqrt{2}\end{equation}} |
| 133 |
CT scanners collect projections around a |
| 134 |
circle rather than a square. The diameter of this circle is |
| 135 |
the diameter of the boundary square \latexonly{$p_d$.} |
| 136 |
\latexignore{\emph{Pd}.} |
| 137 |
\latexonly{These relationships are diagrammed in figure~\ref{phantomgeomfig}.} |
| 138 |
\begin{figure} |
| 139 |
\centerline{\image{8cm;0cm}{scangeometry.eps}} |
| 140 |
\latexonly{\caption{\label{phantomgeomfig} Phantom Geometry}} |
| 141 |
\end{figure} |
| 142 |
|
| 143 |
\subsubsection{View Diameter}\index{View diameter} |
| 144 |
The \emph{view diameter} is the area that is being processed |
| 145 |
during scanning of phantoms as well as during rasterization of |
| 146 |
phantoms. By default, the \emph{view diameter} is set equal |
| 147 |
to the \emph{phantom diameter}. It may be useful, especially for |
| 148 |
experimental reasons, to process an area larger (and maybe even |
| 149 |
smaller) than the phantom. Thus, during rasterization or during |
| 150 |
projections, \ctsim\ will ask for a \emph{view ratio}, |
| 151 |
\latexonly{$v_r$.}\latexignore{\rtfsp \emph{VR}.} The \emph{view |
| 152 |
diameter} is then calculated as |
| 153 |
\latexonly{\begin{equation}v_d = p_dv_r\end{equation}} |
| 154 |
\latexignore{\\\centerline{\emph{Vd = Pd x VR}}\\} |
| 155 |
|
| 156 |
By using a |
| 157 |
\latexonly{$v_r$}\latexignore{\emph{VR}} |
| 158 |
less than 1, \ctsim\ will allow |
| 159 |
for a \emph{view diameter} less than |
| 160 |
\emph{phantom diameter}. |
| 161 |
This will lead to significant artifacts. Physically, this would |
| 162 |
be impossible and is analogous to inserting an object into the CT |
| 163 |
scanner that is larger than the scanner itself! |
| 164 |
|
| 165 |
\subsubsection{Scan Diameter}\index{Scan diameter} |
| 166 |
By default, the entire \emph{view diameter} is scanned. For |
| 167 |
experimental purposes, it may be desirable to scan an area either |
| 168 |
larger or smaller than the \emph{view diameter}. Thus, the concept |
| 169 |
of \emph{scan ratio}, \latexonly{$s_r$,}\latexignore{\emph{SR},} |
| 170 |
arises. The scan diameter, |
| 171 |
\latexonly{$s_d$,}\latexignore{\emph{Sd},} is the diameter over |
| 172 |
which x-rays are collected and is defined as |
| 173 |
\latexonly{\begin{equation}s_d =v_d s_r\end{equation}} |
| 174 |
\latexignore{\\\centerline{\emph{Sd = Vd x SR}}\\} |
| 175 |
By default and |
| 176 |
for all ordinary scanning, the \emph{scan ratio} is to \texttt{1}. |
| 177 |
If the \emph{scan ratio} is less than \texttt{1}, you can expect |
| 178 |
significant artifacts. |
| 179 |
|
| 180 |
\subsubsection{Focal Length}\index{Focal length} |
| 181 |
The \emph{focal length}, |
| 182 |
\latexonly{$f$,}\latexignore{\emph{F},} |
| 183 |
is the distance of the X-ray source to the center of |
| 184 |
the phantom. The focal length is set as a ratio, |
| 185 |
\latexonly{$f_r$,}\latexignore{\emph{FR},} |
| 186 |
of the view radius. Focal length is |
| 187 |
calculated as |
| 188 |
\latexonly{\begin{equation}f = (v_d / 2) f_r\end{equation}} |
| 189 |
\latexignore{\\\centerline{\emph{F = (Vd / 2) x FR}}} |
| 190 |
|
| 191 |
For parallel geometry scanning, the focal length doesn't matter. |
| 192 |
However, for divergent geometry scanning (equilinear and equiangular), |
| 193 |
the \emph{focal length ratio} should be set at \texttt{2} or more |
| 194 |
to avoid artifacts. Moreover, a value of less than \texttt{1} is |
| 195 |
physically impossible and it analagous to having the x-ray |
| 196 |
source inside of the \emph{view diameter}. |
| 197 |
|
| 198 |
\subsubsection{Center-Detector Length}\index{Center-Detector length} |
| 199 |
The \emph{center-detector length}, |
| 200 |
\latexonly{$c$,}\latexignore{\emph{C},} |
| 201 |
is the distance from the center of |
| 202 |
the phantom to the center of the detector array. The center-detector length is set as a ratio, |
| 203 |
\latexonly{$c_r$,}\latexignore{\emph{CR},} |
| 204 |
of the view radius. The center-detector length is |
| 205 |
calculated as |
| 206 |
\latexonly{\begin{equation}f = (v_d / 2) c_r\end{equation}} |
| 207 |
\latexignore{\\\centerline{\emph{F = (Vd / 2) x CR}}} |
| 208 |
|
| 209 |
For parallel geometry scanning, the center-detector length doesn't matter. |
| 210 |
A value of less than \texttt{1} is physically impossible and it analagous to |
| 211 |
having the detector array inside of the \emph{view diameter}. |
| 212 |
|
| 213 |
|
| 214 |
\subsection{Parallel Geometry}\label{geometryparallel}\index{Parallel geometry}\index{Scanner!Parallel} |
| 215 |
|
| 216 |
The simplest geometry, parallel, was used in first generation |
| 217 |
scanners. As mentioned above, the focal length is not used in this simple |
| 218 |
geometry. The detector array is set to be the same size as the |
| 219 |
\emph{scan diameter}. For optimal scanning in this geometry, the |
| 220 |
\emph{scan diameter} should be equal to the \emph{phantom |
| 221 |
diameter}. This is accomplished by using the default values of |
| 222 |
\texttt{1} for the \emph{view ratio} and the \emph{scan ratio}. If |
| 223 |
values of less than \texttt{1} are used for these two variables, |
| 224 |
significant distortions will occur. |
| 225 |
|
| 226 |
|
| 227 |
\subsection{Divergent Geometries}\label{geometrydivergent}\index{Equilinear geometry}\index{Equiangular geometry} |
| 228 |
\index{Scanner!Equilinear}\index{Scanner!Equiangular} |
| 229 |
For both equilinear (second generation) and equiangular |
| 230 |
(third, fourth, and fifth generation) geometries, |
| 231 |
the x-ray beams diverge from a single source to a detector array. |
| 232 |
In the equilinear mode, a single |
| 233 |
source produces a fan beam which is read by a linear array of detectors. If |
| 234 |
the detectors occupy an arc of a circle, then the geometry is equiangular. |
| 235 |
\latexonly{These configurations are shown in figure~\ref{divergentfig}.} |
| 236 |
\begin{figure} |
| 237 |
\centerline{\image{10cm;0cm}{divergent.eps}} |
| 238 |
\latexonly{\caption{\label{divergentfig} Equilinear and equiangular geometries.}} |
| 239 |
\end{figure} |
| 240 |
|
| 241 |
|
| 242 |
\subsubsection{Fan Beam Angle}\index{Fan beam angle} |
| 243 |
For these divergent beam geometries, the \emph{fan beam angle} |
| 244 |
needs to be calculated. For real-world CT scanners, this is fixed |
| 245 |
at the time of manufacture. \ctsim, however, calculates the |
| 246 |
\emph{fan beam angle}, $\alpha$, from the \emph{scan diameter} and |
| 247 |
the \emph{focal length} as |
| 248 |
\latexignore{\centerline{\emph{alpha = 2 x asin ( |
| 249 |
(Sd / 2) / f)}}} |
| 250 |
\latexonly{\begin{equation}\label{alphacalc}\alpha = 2 \sin^{-1} |
| 251 |
((s_d / 2) / f)\end{equation}} |
| 252 |
\latexonly{This is illustrated in figure~\ref{alphacalcfig}.} |
| 253 |
\begin{figure} |
| 254 |
\centerline{\image{10cm;0cm}{alphacalc.eps}} |
| 255 |
\latexonly{\caption{\label{alphacalcfig} Calculation of $\alpha$}} |
| 256 |
\end{figure} |
| 257 |
|
| 258 |
|
| 259 |
Empiric testing with \ctsim\ shows that for very large \emph{fan beam angles}, |
| 260 |
greater than approximately |
| 261 |
\latexonly{$120^\circ$,}\latexignore{120 degrees,} |
| 262 |
there are significant artifacts. The primary way to manage the |
| 263 |
\emph{fan beam angle} is by varying the \emph{focal length} since the |
| 264 |
\emph{scan diameter} is usually fixed at the size of the phantom. |
| 265 |
|
| 266 |
To illustrate, the \emph{scan diameter} can be defined as |
| 267 |
\latexonly{\begin{equation}s_d = s_r v_r p_d\end{equation}} |
| 268 |
\latexignore{\\\centerline{\emph{Sd = Sr x Vr x Pd}}\\} |
| 269 |
|
| 270 |
Further, the \emph{focal length} can be defined as |
| 271 |
\latexonly{\begin{equation} f = f_r (v_r p_d / 2)\end{equation}} |
| 272 |
\latexignore{\\\centerline{\emph{F = FR x (VR x Pd)$$\\}}} |
| 273 |
|
| 274 |
Substituting these equations into \latexignore{the above |
| 275 |
equation,}\latexonly{equation~\ref{alphacalc},} We have, |
| 276 |
\latexonly{ |
| 277 |
\begin{eqnarray} |
| 278 |
\alpha &=& 2\,\sin^{-1} \frac{\displaystyle s_r v_r p_d / 2}{\displaystyle f_r v_r (p_d / 2)} \nonumber \\ |
| 279 |
&=& 2\,\sin^{-1} (s_r / f_r) |
| 280 |
\end{eqnarray} |
| 281 |
} \latexignore{\\\centerline{\emph{\alpha = 2 sin (Sr / Fr)}}\\} |
| 282 |
|
| 283 |
Since in normal scanning $s_r$ = 1, $\alpha$ depends only upon the |
| 284 |
\emph{focal length ratio} in normal scanning. |
| 285 |
|
| 286 |
\subsubsection{Detector Array Size} |
| 287 |
In general, you do not need to be concerned with the detector |
| 288 |
array size -- it is automatically calculated by \ctsim. For the |
| 289 |
particularly interested, this section explains how the detector |
| 290 |
array size is calculated. |
| 291 |
|
| 292 |
For parallel geometry, the detector length is simply the scan |
| 293 |
diameter. |
| 294 |
|
| 295 |
For divergent beam geometries, the size of the detector array also |
| 296 |
depends upon the \emph{focal length}: increasing the \emph{focal |
| 297 |
length} decreases the size of the detector array. |
| 298 |
|
| 299 |
For equiangular geometry, the detectors are equally spaced around a arc |
| 300 |
covering an angular distance of $\alpha$ as viewed from the source. When |
| 301 |
viewed from the center of the scanning, the angular distance is |
| 302 |
\latexonly{$$\pi + \alpha - 2 \, \cos^{-1} \Big( \frac{s_d / 2}{c} \Big)$$} |
| 303 |
\latexignore{\\\emph{pi + \alpha - 2 x acos ((Sd / 2) / C))}\\} |
| 304 |
The dotted circle |
| 305 |
\latexonly{in figure~\ref{equiangularfig}} |
| 306 |
indicates the positions of the detectors in this case. |
| 307 |
|
| 308 |
\begin{figure} |
| 309 |
\centerline{\image{10cm;0cm}{equiangular.eps}} |
| 310 |
\latexonly{\caption{\label{equiangularfig}Equiangular geometry}} |
| 311 |
\end{figure} |
| 312 |
|
| 313 |
For equilinear geometry, the detectors are equally spaced along a straight |
| 314 |
line. The detector length depends upon |
| 315 |
\latexonly{$\alpha$}\latexignore{\emph{alpha}} and the \emph{focal |
| 316 |
length}. This length, |
| 317 |
\latexonly{$d_l$,}\latexignore{Dl,} is calculated as |
| 318 |
\latexonly{\begin{equation} d_l = 2\,(f + c) \tan (\alpha / 2)\end{equation}} |
| 319 |
\latexignore{\\\centerline{\emph{2 x (F + C) x tan(\alpha/2)}}} |
| 320 |
\latexonly{This geometry is shown in figure~\ref{equilinearfig}.} |
| 321 |
\begin{figure} |
| 322 |
\centerline{\image{10cm;0cm}{equilinear.eps}} |
| 323 |
\latexonly{\caption{\label{equilinearfig} Equilinear geometry}} |
| 324 |
\end{figure} |
| 325 |
|
| 326 |
|
| 327 |
\section{Reconstruction}\label{conceptreconstruction}\index{Reconstruction overview}% |
| 328 |
|
| 329 |
\subsection{Direct Inverse Fourier} |
| 330 |
This method is not currently implemented in \ctsim; however, it is |
| 331 |
planned for a future release. This method does not give results as |
| 332 |
accurate as filtered backprojection. This is due primarily |
| 333 |
to interpolation occurring in the frequency domain rather than the |
| 334 |
spatial domain. |
| 335 |
|
| 336 |
\subsection{Filtered Backprojection}\index{Filtered backprojection}\index{Symmetric multiprocessing}\index{SMP} |
| 337 |
The technique is comprised of two sequential steps: |
| 338 |
filtering projections followed by backprojecting the filtered projections. Though |
| 339 |
these two steps are sequential, each view position can be processed independently. |
| 340 |
|
| 341 |
\subsubsection{Parallel Computer Processing}\index{Parallel processing} |
| 342 |
Since each view can be processed independently, filtered backprojection is amendable to |
| 343 |
parallel processing. Indeed, this has been used in commercial scanners to speed reconstruction. |
| 344 |
This parallelism is exploited both in the \ctsim\ graphical shell and |
| 345 |
in the \helpref{LAM}{ctsimtextlam} version of \ctsimtext. \ctsim\ can distribute it's workload |
| 346 |
amongst multiple processors working in parallel. |
| 347 |
|
| 348 |
The graphical shell will automatically take advantage of multiple CPU's when |
| 349 |
running on a \emph{Symmetric Multiprocessing} |
| 350 |
computer. Dual-CPU computers are commonly available which provide a near doubling |
| 351 |
in reconstruction speeds. \ctsim, though, has no limits on the number of CPU's |
| 352 |
that can be used in parallel. The \emph{LAM} version |
| 353 |
of \ctsimtext\ is designed to work in a cluster of computers. |
| 354 |
This has been testing with a cluster of 16 computers in a |
| 355 |
\urlref{Beowulf-class}{http://www.beowulf.org} cluster with excellent |
| 356 |
results. |
| 357 |
|
| 358 |
\subsubsection{Filter projections} |
| 359 |
The first step in filtered backprojection reconstructions is the filtering |
| 360 |
of each projection. The projections for a each view have their frequency data multipled by |
| 361 |
a filter of $|w|$. \ctsim\ permits four different ways to accomplish this |
| 362 |
filtering. |
| 363 |
|
| 364 |
Two of the methods use convolution of the projection data with the |
| 365 |
inverse Fourier transform of $|w|$. The other two methods perform an Fourier |
| 366 |
transform of the projection data and multiply that by the $|w|$ filter and |
| 367 |
then perform an inverse fourier transform. |
| 368 |
|
| 369 |
Though multiplying by $|w|$ gives the sharpest reconstructions, in |
| 370 |
practice, superior results are obtained by reducing the higher |
| 371 |
frequencies. This is performed by mutiplying the $|w|$ filter by |
| 372 |
another filter that attenuates the higher frequencies. \ctsim\ has |
| 373 |
multiple filters for this purpose. |
| 374 |
|
| 375 |
\subsubsection{Backprojection of filtered projections} |
| 376 |
Backprojection is the process of ``smearing'' the filtered |
| 377 |
projections over the reconstructing image. Various levels of |
| 378 |
interpolation can be specified. |
| 379 |
|
| 380 |
\section{Image Comparison}\label{conceptimagecompare}\index{Image!Comparison} |
| 381 |
Images can be compared statistically. Three measurements can be calculated |
| 382 |
by \ctsim. They are taken from the standard measurements used by |
| 383 |
Herman\cite{HERMAN80}. They are: |
| 384 |
|
| 385 |
\begin{itemize}\itemsep=0pt |
| 386 |
\item[]\textbf{$d$}\quad The normalized root mean squared distance measure. |
| 387 |
\item[]\textbf{$r$}\quad The normalized mean absolute distance measure. |
| 388 |
\item[]\textbf{$e$}\quad The worst case distance measure over a \latexonly{$2\times2$}\latexignore{\emph{2 x 2}} pixel area. |
| 389 |
\end{itemize} |
| 390 |
|
| 391 |
These measurements are defined in equations \ref{dequation} through \ref{bigrequation}. |
| 392 |
In these equations, $p$ denotes the phantom image, $r$ denotes the reconstruction |
| 393 |
image, and $\bar{p}$ denotes the average pixel value of $p$. Each of the images have a |
| 394 |
size of $m \times n$. In equation \ref{eequation} $[n/2]$ and $[m/2]$ denote the largest |
| 395 |
integers less than $n/2$ and $m/2$, respectively. |
| 396 |
|
| 397 |
\latexignore{These formulas are shown in the print documentation of \ctsim.} |
| 398 |
% |
| 399 |
%Tex2RTF can not handle the any subscripts or superscripts for the inner summation unless |
| 400 |
% have a space character before the \sum |
| 401 |
\latexonly{\begin{equation}\label{dequation} d =\sqrt{\frac{\displaystyle \sum_{i=1}^{n}{ \sum_{j=1}^{m}{(p_{i,j} - r_{i,j})^2}}}{\displaystyle \sum_{i=1}^{n}{ \sum_{j=1}^{m}{(p_{i,j} - \bar{p})^2}}}}\end{equation}} |
| 402 |
\latexonly{\begin{equation}\label{requation}r = \frac{ \displaystyle \sum_{i=1}^{n}{ \sum_{j=1}^{m}{|p_{i,j} - r_{i,j}|}}}{ \displaystyle \sum_{i=1}^{n}{ \sum_{j=1}^{m}{|p_{i,j}|}}}\end{equation}} |
| 403 |
\latexonly{\begin{equation}\label{eequation}e = \max_{1 \le k \le [n/2] \atop 1 \le l \le [m/2]}(|P_{k,l} - R_{k,l}|)\end{equation}} |
| 404 |
\latexonly{where} |
| 405 |
\latexonly{\begin{equation}\label{bigpequation}P_{k,l} = \textstyle \frac{1}{4} (p_{2k,2l} + p_{2k+1,2l} + p_{2k,2l+l} + p_{2k+1,2l+1})\end{equation}} |
| 406 |
\latexonly{\begin{equation}\label{bigrequation}R_{k,l} = \textstyle \frac{1}{4} (r_{2k,2l} + r_{2k+1,2l} + r_{2k,2l+1} + r_{2k+1,2l+1})\end{equation}} |
| 407 |
\begin{comment} |
| 408 |
\end{comment} |