lib/PDL/Image2D.pd

filter: Statementstmt 100% 8 / 8 Branchbran 8.3% 92 / 1104 Conditioncond n/a MC/DCmcdc n/a Subroutinesub n/a Podpod n/a totaltotal 8.9% 100 / 1112 CC 0 SCAR 0
CC
0
Cov
0%
CRAP 0
« lib/PDL/ImageND-pp-convolveND.c ↑ Summary lib/PDL/MatrixOps-pp-squaretotri.c »
1 use strict;
2 use warnings;
3 use PDL::Types qw(ppdefs_all);
4 my $A = [ppdefs_all];
5  
6 { no warnings 'once'; # pass info back to Makefile.PL
7 $PDL::Core::Dev::EXTRAS{$::PDLMOD}{OBJECT} .= join '', map " $::PDLBASE/$_\$(OBJ_EXT)", qw(rotate resample select misc);
8 $PDL::Core::Dev::EXTRAS{$::PDLMOD}{INC} .= qq{ "-I$::PDLBASE"};
9 }
10  
11 pp_addpm({At=>'Top'},<<'EOD');
12 use strict;
13 use warnings;
14  
15 =head1 NAME
16  
17 PDL::Image2D - Miscellaneous 2D image processing functions
18  
19 =head1 DESCRIPTION
20  
21 Miscellaneous 2D image processing functions - for want
22 of anywhere else to put them.
23  
24 =head1 SYNOPSIS
25  
26 use PDL::Image2D;
27  
28 =cut
29  
30 use PDL; # ensure qsort routine available
31 use PDL::Math;
32 use Carp;
33  
34 my %boundary2value = (Reflect=>1, Truncate=>2, Replicate=>3);
35 EOD
36  
37 pp_addpm({At=>'Bot'},<<'EOD');
38  
39 =head1 AUTHORS
40  
41 Copyright (C) Karl Glazebrook 1997 with additions by Robin Williams
42 (rjrw@ast.leeds.ac.uk), Tim Jenness (timj@jach.hawaii.edu),
43 and Doug Burke (burke@ifa.hawaii.edu).
44  
45 All rights reserved. There is no warranty. You are allowed
46 to redistribute this software / documentation under certain
47 conditions. For details, see the file COPYING in the PDL
48 distribution. If this file is separated from the PDL distribution,
49 the copyright notice should be included in the file.
50  
51 =cut
52  
53 EOD
54  
55 #################################################
56 # BEGIN INTERNAL FUNCTION DECLARATIONS #
57 #################################################
58  
59 pp_def('polyfill_pp',
60 HandleBad => 0, # a marker
61 Pars => '[io] im(m,n); float ps(two=2,np); int col()',
62 GenericTypes => ['L'],
63 CHeader => qq{void polyfill(PDL_Long *image, int wx, int wy, float *ps, int n, PDL_Long col, int *ierr);\n},
64 Code => 'int ierr = 0, nerr;
65 broadcastloop %{
66 polyfill($P(im), $SIZE(m), $SIZE(n), $P(ps), $SIZE(np), $col(), &nerr);
67 ierr = PDLMAX(ierr, nerr);
68 %}
69 if (ierr) warn("errors during polygonfilling");
70 ',
71 Doc => undef,
72 PMFunc => ''
73 );
74  
75 my %pnpolyFields = (
76 'pnpoly_pp' => {'pars' => 'a(m,n); ps(k,l); int [o] msk(m,n)', 'special' => '$msk() = c;'},
77 'pnpolyfill_pp' => {'pars' => '[io] a(m,n); ps(k,l); int col()', 'special' => 'if(c) { $a() = $col(); }'}
78 );
79  
80 for my $name (sort keys %pnpolyFields) {
81 pp_def($name,
82 HandleBad => 0,
83 PMFunc => '',
84 Doc => undef,
85 Pars => $pnpolyFields{$name}->{'pars'},
86 Code => '
87 #define VERTX(q) $ps(k=>0,l=>q)
88 #define VERTY(q) $ps(k=>1,l=>q)
89  
90 broadcastloop %{
91 loop(n,m) %{
92 PDL_Indx j = $SIZE(l)-1, c = 0;
93 loop (l) %{
94 if ( ((VERTY(l)>n) != (VERTY(j)>n)) &&
95 (m < (VERTX(j)-VERTX(l)) * (n-VERTY(l)) / (VERTY(j)-VERTY(l)) + VERTX(l)) )
96 c = !c;
97 j = l;
98 %}
99 ' . $pnpolyFields{$name}{special} .'
100 %}
101 %}
102  
103 #undef VERTX
104 #undef VERTY
105 '
106 );
107 }
108  
109 pp_export_nothing(); # Clear the export list
110  
111 #################################################
112 # END INTERNAL FUNCTION DECLARATIONS #
113 #################################################
114  
115  
116 pp_addhdr('
117  
118 /* Fast Modulus with proper negative behaviour */
119  
120 #define REALMOD(a,b) {while ((a)>=(b)) (a) -= (b); while ((a)<0) (a) += (b);}
121  
122 #define X(symbol, ctype, ppsym, ...) \
123 ctype quick_select_ ## ppsym(ctype arr[], int n);
124 PDL_TYPELIST_REAL(X)
125 #undef X
126 ');
127  
128 my %init =
129 (
130 i => { size => 'm_size', off => 'poff', init => '1-p_size' },
131 j => { size => 'n_size', off => 'qoff', init => '1-q_size' },
132 );
133  
134 # requires 'int $var, ${var}2' to have been declared in the c code
135 # (along with [pq]off and [pq]_size)
136 #
137 sub init_map {
138 my $var = shift;
139  
140 my $loop = $var;
141 my $loop2 = "${var}2";
142  
143 my $href = $init{$var} ||
144 die "ERROR: unknown variable sent to init_map()\n";
145 my $size = $href->{size} ||
146 die "ERROR: unable to find size for $var\n";
147 my $off = $href->{off} ||
148 die "ERROR: unable to find off for $var\n";
149 my $init = $href->{init} ||
150 die "ERROR: unable to find init for $var\n";
151  
152 return
153 "for ( $loop = $init; $loop< $size; ${loop}++) {
154 $loop2 = $loop + $off;
155 switch (opt) {
156 case 1: /* REFLECT */
157 if (${loop2}<0)
158 $loop2 = -${loop2}-1;
159 else if ($loop2 >= $size)
160 $loop2 = 2*${size}-(${loop2}+1);
161 break;
162 case 2: /* TRUNCATE */
163 if (${loop2}<0 || ${loop2} >= $size)
164 $loop2 = -1;
165 break;
166 case 3: /* REPLICATE */
167 if (${loop2}<0)
168 $loop2 = 0;
169 if (${loop2} >= $size)
170 $loop2 = $size-1;
171 break;
172 default:
173 REALMOD($loop2,$size);
174 }
175 map${var}\[$loop] = $loop2;
176 }\n";
177  
178 } # sub: init_map()
179  
180 sub init_vars {
181 my $sizevars = shift || [qw(m n p q)];
182 my $compvars = shift || [];
183 my $str = "int i,j, i2,j2, poff, qoff;";
184 $str .=
185 'int opt = $COMP(opt);
186 ' . join("\n", map "int ${_}_size = \$SIZE(${_});", @$sizevars) . '
187 ' . join("\n", map "int ${_}_size = \$COMP(${_}_size);", @$compvars) . '
188 PDL_Indx *mapi = $P(mapi), *mapj = $P(mapj);
189 if ((mapi==NULL) || (mapj==NULL)) $CROAK("Out of Memory");
190  
191 poff = p_size/2; mapi += p_size-1;
192 qoff = q_size/2; mapj += q_size-1;
193 ';
194  
195 return $str;
196 } # sub: init_vars()
197  
198 pp_def('conv2d', Doc=><<'EOD',
199 =for ref
200  
201 2D convolution of an array with a kernel (smoothing)
202  
203 For large kernels, using a FFT routine,
204 such as L<PDL::FFT/fftconvolve>,
205 will be quicker.
206  
207 =for usage
208  
209 $new = conv2d $old, $kernel, {OPTIONS}
210  
211 =for example
212  
213 $smoothed = conv2d $image, ones(3,3), {Boundary => Reflect}
214  
215 =for options
216  
217 Boundary - controls what values are assumed for the image when kernel
218 crosses its edge:
219 => Default - periodic boundary conditions
220 (i.e. wrap around axis)
221 => Reflect - reflect at boundary
222 => Truncate - truncate at boundary
223 => Replicate - repeat boundary pixel values
224  
225 =cut
226 EOD
227 BadDoc =>
228 'Unlike the FFT routines, conv2d is able to process bad values.',
229 HandleBad => 1,
230 Pars => 'a(m,n); kern(p,q); [o]b(m,n); indx [t]mapi(isize=CALC($SIZE(p) + $SIZE(m))); indx [t]mapj(jsize=CALC($SIZE(q) + $SIZE(n)))',
231 OtherPars => 'int opt;',
232 PMCode => '
233 sub PDL::conv2d {
234 my $opt; $opt = pop @_ if ref($_[$#_]) eq \'HASH\';
235 die \'Usage: conv2d( a(m,n), kern(p,q), [o]b(m,n), {Options} )\'
236 if $#_<1 || $#_>2;
237 my($x,$kern) = @_;
238 my $c = $#_ == 2 ? $_[2] : $x->nullcreate;
239 PDL::_conv2d_int($x,$kern,$c,
240 (!($opt && exists $$opt{Boundary}))?0:$boundary2value{$$opt{Boundary}}
241 );
242 return $c;
243 }
244 ',
245 GenericTypes => $A,
246 Code =>
247 init_vars() .
248 init_map("i") .
249 init_map("j") .
250 pp_line_numbers(__LINE__, <<'EOF'),
251 broadcastloop %{
252 loop(n,m) %{
253 PDL_IF_BAD(int flag = 0;,)
254 PDL_CLDouble tmp = 0;
255 loop (q) %{
256 j2 = mapj[n-q];
257 if (j2 >= 0) {
258 loop (p) %{
259 i2 = mapi[m-p];
260 if (i2 >= 0) {
261 PDL_IF_BAD(if ( $ISGOOD(a(m=>i2,n=>j2)) && $ISGOOD(kern()) ),) {
262 tmp += $a(m=>i2,n=>j2) * $kern();
263 PDL_IF_BAD(flag = 1;,)
264 } /* if: good */
265 } /* if: i2 >= 0 */
266 %}
267 } /* if: j2 >= 0 */
268 %}
269 PDL_IF_BAD(if ( !flag ) { $SETBAD(b()); }
270 else,) { $b() = tmp; }
271 %}
272 %}
273 EOF
274 ); # pp_def: conv2d
275  
276 pp_def('med2d', Doc=> <<'EOD',
277 =for ref
278  
279 2D median-convolution of an array with a kernel (smoothing)
280  
281 Note: only points in the kernel E<gt>0 are included in the median, other
282 points are weighted by the kernel value (medianing lots of zeroes
283 is rather pointless)
284  
285 =for usage
286  
287 $new = med2d $old, $kernel, {OPTIONS}
288  
289 =for example
290  
291 $smoothed = med2d $image, ones(3,3), {Boundary => Reflect}
292  
293 =for options
294  
295 Boundary - controls what values are assumed for the image when kernel
296 crosses its edge:
297 => Default - periodic boundary conditions (i.e. wrap around axis)
298 => Reflect - reflect at boundary
299 => Truncate - truncate at boundary
300 => Replicate - repeat boundary pixel values
301  
302 =cut
303 EOD
304 BadDoc =>
305 'Bad values are ignored in the calculation. If all elements within the
306 kernel are bad, the output is set bad.',
307 HandleBad => 1,
308 Pars => 'a(m,n); kern(p,q); [o]b(m,n); double+ [t]tmp(pq=CALC($SIZE(p)*$SIZE(q))); indx [t]mapi(isize=CALC($SIZE(p) + $SIZE(m))); indx [t]mapj(jsize=CALC($SIZE(q) + $SIZE(n)))',
309 OtherPars => 'int opt;',
310 PMCode => pp_line_numbers(__LINE__, <<'EOF'),
311 sub PDL::med2d {
312 my $opt; $opt = pop @_ if ref($_[$#_]) eq 'HASH';
313 die 'Usage: med2d( a(m,n), kern(p,q), [o]b(m,n), {Options} )'
314 if $#_<1 || $#_>2;
315 my($x,$kern) = @_;
316 croak "med2d: kernel must contain some positive elements.\n"
317 if all( $kern <= 0 );
318 my $c = $#_ == 2 ? $_[2] : $x->nullcreate;
319 PDL::_med2d_int($x,$kern,$c,
320 (!($opt && exists $$opt{Boundary}))?0:$boundary2value{$$opt{Boundary}}
321 );
322 return $c;
323 }
324 EOF
325 CHeader => 'PDL_TYPELIST_FLOATREAL(PDL_QSORT)',
326 Code =>
327 init_vars() .
328 init_map("i") .
329 init_map("j") .
330 pp_line_numbers(__LINE__, <<'EOF'),
331 broadcastloop %{
332 loop (n,m) %{
333 PDL_Indx count = 0;
334 PDL_IF_BAD(int flag = 0;,)
335 loop (q) %{
336 j2 = mapj[n-q];
337 if (j2 >= 0)
338 loop (p) %{
339 i2 = mapi[m-p];
340 if (i2 >= 0) {
341 PDL_LDouble kk = $kern(), aa = $a(m=>i2,n=>j2);
342 PDL_IF_BAD(if ( $ISGOODVAR(kk,kern) && $ISGOODVAR(aa,a) ),) {
343 PDL_IF_BAD(flag = 1;,)
344 if ( kk > 0 )
345 $tmp(pq=>count++) = aa * kk;
346 }
347 } /* if: i2 >= 0 */
348 %}
349 %}
350 PDL_IF_BAD(if ( flag == 0 ) {
351 $SETBAD(b());
352 } else,) {
353 qsort_$PPSYM(tmp)( $P(tmp), 0, count-1 );
354 $b() = $tmp(pq=>(count-1)/2);
355 }
356 %}
357 %}
358 EOF
359 ); # pp_def: med2d
360  
361 pp_def('med2df', Doc=> <<'EOD',
362  
363 =for ref
364  
365 2D median-convolution of an array in a pxq window (smoothing)
366  
367 Note: this routine does the median over all points in a rectangular
368 window and is not quite as flexible as C<med2d> in this regard
369 but slightly faster instead
370  
371 =for usage
372  
373 $new = med2df $old, $xwidth, $ywidth, {OPTIONS}
374  
375 =for example
376  
377 $smoothed = med2df $image, 3, 3, {Boundary => Reflect}
378  
379 =for options
380  
381 Boundary - controls what values are assumed for the image when kernel
382 crosses its edge:
383 => Default - periodic boundary conditions (i.e. wrap around axis)
384 => Reflect - reflect at boundary
385 => Truncate - truncate at boundary
386 => Replicate - repeat boundary pixel values
387  
388 =cut
389  
390 EOD
391 Pars => 'a(m,n); [o]b(m,n); indx [t]mapi(isize=CALC($SIZE(p) + $SIZE(m))); indx [t]mapj(jsize=CALC($SIZE(q) + $SIZE(n)))',
392 OtherPars => 'IV p_size=>p; IV q_size=>q; int opt;',
393 PMCode => <<'EOF',
394 sub PDL::med2df {
395 my $opt; $opt = pop @_ if ref($_[$#_]) eq 'HASH';
396 die 'Usage: med2df( a(m,n), [o]b(m,n), p, q, {Options} )'
397 if $#_<2 || $#_>3;
398 my($x,$p,$q) = @_;
399 croak "med2df: kernel must contain some positive elements.\n"
400 if $p == 0 && $q == 0;
401 my $c = $#_ == 3 ? $_[3] : $x->nullcreate;
402 &PDL::_med2df_int($x,$c,$p,$q,
403 (!($opt && exists $$opt{Boundary}))?0:$boundary2value{$$opt{Boundary}}
404 );
405 return $c;
406 }
407 EOF
408 Code =>
409 init_vars() .
410 init_map("i") .
411 init_map("j") .
412 '
413 $GENERIC() tmp[p_size*q_size];
414 broadcastloop %{
415 loop (n,m) %{
416 int count = 0;
417 loop (q) %{
418 if ((j2 = mapj[n-q]) < 0) continue;
419 loop (p) %{
420 if ((i2 = mapi[m-p]) < 0) continue;
421 tmp[count++] = $a(m=>i2,n=>j2);
422 %}
423 %}
424 $b() = quick_select_$PPSYM() (tmp, count );
425 %}
426 %}
427 ',
428  
429 ); # pp_def: med2df
430  
431 pp_addhdr(<<'EOH');
432 #define EZ(x) ez ? 0 : (x)
433 EOH
434 pp_def('box2d',
435 Pars => 'a(n,m); [o] b(n,m)',
436 OtherPars => 'int wx; int wy; int edgezero',
437 GenericTypes => $A,
438 Code => '
439 register int nx = 0.5*$COMP(wx);
440 register int ny = 0.5*$COMP(wy);
441 register int ez = $COMP(edgezero);
442 PDL_CLDouble div, sum, lsum;
443 int ind1,ind2,first;
444  
445 div = 1/((2.0*nx+1)*(2.0*ny+1));
446  
447 broadcastloop %{
448 first = 1;
449 loop (m,n=:nx) %{
450 ind1 = $SIZE(n)-1-n; /* rightmost strip */
451 $b() = EZ($a());
452 $b(n=>ind1) = EZ($a(n=>ind1));
453 %}
454 loop (n,m=:ny) %{
455 ind1 = $SIZE(m)-1-m; /* topmost strip */
456 $b() = EZ($a());
457 $b(m=>ind1) = EZ($a(m=>ind1));
458 %}
459 loop (m=ny:-ny) %{
460 if (first) {
461 first = 0;
462 lsum = 0;
463 PDL_Indx m_outer = m;
464 loop (m=m_outer-ny:m_outer+ny+1,n=:2*nx+1) %{
465 lsum += $a(); /* initialize sum */
466 %}
467 } else {
468 ind1 = m-ny-1;
469 ind2 = m+ny;
470 loop (n=:2*nx+1) %{
471 lsum -= $a(m=>ind1); /* remove top pixels */
472 lsum += $a(m=>ind2); /* add bottom pixels */
473 %}
474 }
475 sum = lsum;
476 $b(n=>nx) = div*sum; /* and assign */
477 loop (n=nx+1:-nx) %{ /* loop along line */
478 ind1 = n-nx-1;
479 ind2 = n+nx;
480 PDL_Indx m_outer = m;
481 loop (m=m_outer-ny:m_outer+ny+1) %{
482 sum -= $a(n=>ind1); /* remove leftmost data */
483 sum += $a(n=>ind2); /* and add rightmost */
484 %}
485 $b() = div*sum; /* and assign */
486 %}
487 %}
488 %}',
489 Doc => << 'EOD',
490 =for ref
491  
492 fast 2D boxcar average
493  
494 =for example
495  
496 $smoothim = $im->box2d($wx,$wy,$edgezero=1);
497  
498 The edgezero argument controls if edge is set to zero (edgezero=1)
499 or just keeps the original (unfiltered) values.
500  
501 C<box2d> should be updated to support similar edge options
502 as C<conv2d> and C<med2d> etc.
503  
504 Boxcar averaging is a pretty crude way of filtering. For serious stuff
505 better filters are around (e.g., use L</conv2d> with the appropriate
506 kernel). On the other hand it is fast and computational cost grows only
507 approximately linearly with window size.
508  
509 =cut
510  
511 EOD
512 ); # pp_def box2d
513  
514 =head2 patch2d
515  
516 =cut
517  
518 pp_def('patch2d',
519 Doc=><<'EOD',
520  
521 =for ref
522  
523 patch bad pixels out of 2D images using a mask
524  
525 C<$bad> is a 2D mask array where 1=bad pixel 0=good pixel.
526 Pixels are replaced by the average of their non-bad neighbours;
527 if all neighbours are bad, the original data value is
528 copied across.
529  
530 =cut
531  
532 EOD
533 BadDoc =>
534 'This routine does not handle bad values - use L</patchbad2d> instead',
535 HandleBad => 0,
536 Pars => 'a(m,n); int bad(m,n); [o]b(m,n);',
537 GenericTypes => $A,
538 Code =>
539 'PDL_Indx i1,j1, i2,j2, norm;
540 PDL_CLDouble tmp;
541 broadcastloop %{
542 loop (n,m) %{
543 $b() = $a();
544 if (!$bad()) continue;
545 tmp = 0; norm=0;
546 for(j1=-1; j1<=1; j1++) {
547 j2 = n+j1;
548 if ( j2<0 || j2>=$SIZE(n) ) continue;
549 for(i1=-1; i1<=1; i1++) {
550 /* ignore central pixel, which we know is bad */
551 if ( i1==0 && j1==0 ) continue;
552 i2 = m+i1;
553 if ( i2<0 || i2>=$SIZE(m) || $bad(m=>i2,n=>j2) ) continue;
554 tmp += $a(m=>i2,n=>j2);
555 norm++;
556 } /* for: i1 */
557 } /* for: j1 */
558 if (norm>0) { /* Patch */
559 $b() = tmp/norm;
560 }
561 %}
562 %} /* broadcastloop */
563 ', # Code
564 );
565  
566 pp_def('patchbad2d',
567 Doc=><<'EOD',
568 =for ref
569  
570 patch bad pixels out of 2D images containing bad values
571  
572 Pixels are replaced by the average of their non-bad neighbours;
573 if all neighbours are bad, the output is set bad.
574 If the input ndarray contains I<no> bad values, then a straight copy
575 is performed (see L</patch2d>).
576 EOD
577 BadDoc =>
578 'patchbad2d handles bad values. The output ndarray I<may> contain
579 bad values, depending on the pattern of bad values in the input ndarray.',
580 HandleBad => 1,
581 Pars => 'a(m,n); [o]b(m,n);',
582 GenericTypes => $A,
583 Code => pp_line_numbers(__LINE__, <<'EOF'),
584 2 PDL_IF_BAD(int flag = 0;,)
585 9 ▶broadcastloop %{
Branch: 210 branches
BranchTrueFalseTotal
broadcastloop %{ 0 2 2
broadcastloop %{ 0 2 2
broadcastloop %{ 0 2 2
broadcastloop %{ 0 2 2
broadcastloop %{ 2 0 2
broadcastloop %{ 1 1 2
broadcastloop %{ 0 2 2
broadcastloop %{ 1 1 2
broadcastloop %{ 1 0 1
broadcastloop %{ 0 2 2
broadcastloop %{ 1 1 2
broadcastloop %{ 0 1 1
broadcastloop %{ 1 1 2
broadcastloop %{ 0 2 2
broadcastloop %{ 0 2 2
broadcastloop %{ 1 1 2
broadcastloop %{ 1 0 1
broadcastloop %{ 0 9 9
broadcastloop %{ 7 2 9
broadcastloop %{ 0 9 9
broadcastloop %{ 0 9 9
broadcastloop %{ 5 4 9
broadcastloop %{ 0 4 4
broadcastloop %{ 0 9 9
broadcastloop %{ 5 4 9
broadcastloop %{ 5 4 9
broadcastloop %{ 11 22 33
broadcastloop %{ 11 22 33
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 1 1 2
broadcastloop %{ 1 1 2
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 0 1 1
broadcastloop %{ 1 1 2
broadcastloop %{ 1 1 2
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
broadcastloop %{ 0 0 0
586 63 ▶ loop(n,m) %{
Branch: 60 branches
BranchTrueFalseTotal
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 25 5 30
loop(n,m) %{ 5 1 6
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 25 5 30
loop(n,m) %{ 5 1 6
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
loop(n,m) %{ 0 0 0
587 51 $GENERIC(a) a_val = $a();
588 134 ▶ PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) {
Branch: 207 branches
BranchTrueFalseTotal
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 25 25
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 9 16 25
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 27 0 27
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 27 27
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 27 54 81
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 9 18 27
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 72 0 72
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 72 72
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 72 72
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 40 32 72
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 81 27 108
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 27 9 36
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 8 1 9
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
PDL_IF_BAD(if ( !$ISGOODVAR(a_val,a) ) { 0 0 0
589 PDL_CLDouble tmp = 0; PDL_Indx norm=0; int i1; int j1;
590 for(j1=-1; j1<=1; j1++) {
591 PDL_Indx j2 = n+j1;
592 if ( j2<0 || j2>=$SIZE(n) ) continue;
593 for(i1=-1; i1<=1; i1++) {
594 /* ignore central pixel, which we know is bad */
595 if ( i1==0 && j1==0 ) continue;
596 PDL_Indx i2 = m+i1;
597 if ( i2<0 || i2>=$SIZE(m) ) continue;
598 a_val = $a(m=>i2,n=>j2);
599 if ( $ISBADVAR(a_val,a) ) continue;
600 tmp += a_val;
601 norm++;
602 } /* for: i1 */
603 } /* for: j1 */
604 /* Patch */
605 if (norm>0) {
606 $b() = tmp/norm;
607 } else {
608 $SETBAD(b());
609 flag = 1;
610 }
611 } else,) {
612 42 $b() = a_val;
613 } /* if: ISGOODVAR() */
614 %}
615 3 ▶%} /* broadcastloop */
Branch: 60 branches
BranchTrueFalseTotal
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 1 1
%} /* broadcastloop */ 0 1 1
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 1 1
%} /* broadcastloop */ 0 1 1
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
%} /* broadcastloop */ 0 0 0
616 /* handle bad flag */
617 2 ▶PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,)
Branch: 15 branches
BranchTrueFalseTotal
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 1 0 1
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
PDL_IF_BAD(if ( flag ) $PDLSTATESETBAD(b);,) 0 0 0
618 EOF
619 );
620  
621 pp_def('max2d_ind',
622 Doc=><<'EOD',
623  
624 =for ref
625  
626 Return value/position of maximum value in 2D image
627  
628 Contributed by Tim Jenness
629  
630 =cut
631  
632 EOD
633  
634 BadDoc=><<'EOD',
635  
636 Bad values are excluded from the search. If all pixels
637 are bad then the output is set bad.
638  
639 EOD
640  
641 HandleBad => 1,
642 Pars => 'a(m,n); [o]val(); int [o]x(); int[o]y();',
643 Code => '
644 PDL_LDouble cur = 0;
645 PDL_Indx curind1 = PDL_IF_BAD(-1,0), curind2 = PDL_IF_BAD(-1,0);
646 loop(m,n) %{
647 if (!(PDL_IF_BAD($ISGOOD(a()) &&,) ( (!n && !m) || $a() > cur || isnan(cur)))) continue;
648 cur = $a(); curind1 = m; curind2 = n;
649 %}
650 PDL_IF_BAD(if ( curind1 < 0 ) {
651 $SETBAD(val());
652 $SETBAD(x());
653 $SETBAD(y());
654 } else,) {
655 $val() = cur;
656 $x() = curind1;
657 $y() = curind2;
658 }
659 ');
660  
661 pp_def('centroid2d',
662 Doc=><<'EOD',
663 =for ref
664  
665 Refine a list of object positions in 2D image by centroiding in a box
666  
667 C<$box> is the full-width of the box, i.e. the window
668 is C<+/- $box/2>.
669  
670 =cut
671 EOD
672 BadDoc=><<'EOD',
673 Bad pixels are excluded from the centroid calculation. If all elements are
674 bad (or the pixel sum is 0 - but why would you be centroiding
675 something with negatives in...) then the output values are set bad.
676 EOD
677 HandleBad => 1,
678 Pars => 'im(m,n); x(); y(); box(); [o]xcen(); [o]ycen();',
679 Code => '
680 PDL_LDouble sum = 0, sumx = 0, sumy = 0;
681 PDL_Indx i1 = $x() - $box()/2;
682 PDL_Indx i2 = $x() + $box()/2 + 1;
683 PDL_Indx j1 = $y() - $box()/2;
684 PDL_Indx j2 = $y() + $box()/2 + 1;
685 loop (n=j1:j2,m=i1:i2) %{
686 PDL_LDouble data = $im();
687 PDL_IF_BAD(if ( $ISBADVAR(data,im) ) continue;,)
688 sum += data;
689 sumx += data*m;
690 sumy += data*n;
691 %}
692 /*
693 * if sum == 0 then we will flag as bad -- although it could just mean that
694 * there is negative values in the dataset.
695 * - should use a better check than != 0.0 ...
696 */
697 PDL_IF_BAD(if ( sum == 0.0 ) {
698 $SETBAD(xcen());
699 $SETBAD(ycen());
700 } else,) {
701 $xcen() = sumx/sum;
702 $ycen() = sumy/sum;
703 }
704 '
705 );
706  
707 pp_add_exported('', 'crop');
708 pp_addpm(<<'EOPM');
709 =head2 crop
710  
711 =for ref
712  
713 Return bounding box of given mask in an C<indx> ndarray, so it can broadcast.
714 Use other operations (such as L<PDL::Bad/isgood>, or
715 L<PDL::Primitive/eqvec> with a colour vector) to create a mask suitable
716 for your application.
717  
718 =for example
719  
720 $x1x2y1y2 = crop($image);
721  
722 =cut
723  
724 *crop = \&PDL::crop;
725 sub PDL::crop {
726 my ($mask) = @_;
727 $mask->xchg(0,1)->orover->_which_int(my $out = null, null);
728 $out->badflag(1); $out->badvalue(-1);
729 my ($x1, $x2) = $out->minmaximum;
730 $mask->orover->_which_int($out = null, null);
731 $out->badflag(1); $out->badvalue(-1);
732 my ($y1, $y2) = $out->minmaximum;
733 $x1->cat($x2, $y1, $y2)->mv(-1,0);
734 }
735 EOPM
736  
737 pp_add_exported('', 'cc8compt','cc4compt');
738 pp_addpm(<<'EOPM');
739  
740 =head2 cc8compt
741  
742 =for ref
743  
744 Connected 8-component labeling of a binary image.
745  
746 Connected 8-component labeling of 0,1 image - i.e. find separate
747 segmented objects and fill object pixels with object number.
748 8-component labeling includes all neighboring pixels.
749 This is just a front-end to ccNcompt. See also L</cc4compt>.
750  
751 =for example
752  
753 $segmented = cc8compt( $image > $threshold );
754  
755 =head2 cc4compt
756  
757 =for ref
758  
759 Connected 4-component labeling of a binary image.
760  
761 Connected 4-component labeling of 0,1 image - i.e. find separate
762 segmented objects and fill object pixels with object number.
763 4-component labling does not include the diagonal neighbors.
764 This is just a front-end to ccNcompt. See also L</cc8compt>.
765  
766 =for example
767  
768 $segmented = cc4compt( $image > $threshold );
769  
770 =cut
771  
772 sub PDL::cc8compt{
773 return ccNcompt(shift,8);
774 }
775 *cc8compt = \&PDL::cc8compt;
776  
777 sub PDL::cc4compt{
778 return ccNcompt(shift,4);
779 }
780 *cc4compt = \&PDL::cc4compt;
781  
782 EOPM
783  
784 pp_def('ccNcompt',Doc=>'
785  
786 =for ref
787  
788 Connected component labeling of a binary image.
789  
790 Connected component labeling of 0,1 image - i.e. find separate
791 segmented objects and fill object pixels with object number.
792 See also L</cc4compt> and L</cc8compt>.
793  
794 The connectivity parameter must be 4 or 8.
795  
796 =for example
797  
798 $segmented = ccNcompt( $image > $threshold, 4);
799  
800 $segmented2 = ccNcompt( $image > $threshold, 8);
801  
802 where the second parameter specifies the connectivity (4 or 8) of the labeling.
803  
804 =cut
805  
806 ',
807 HandleBad => 0, # a marker
808 Pars => 'a(m,n); int+ [o]b(m,n);',
809 OtherPars => 'int con',
810 CHeader => qq{void AddEquiv(PDL_Long* equiv, PDL_Long i, PDL_Long j);\n},
811 Code => '
812  
813 PDL_Long i,k;
814 PDL_Long newlabel;
815 PDL_Long neighbour[4];
816 PDL_Long nfound;
817 PDL_Long pass,count,next,this;
818 PDL_Long *equiv = NULL;
819 PDL_Long i1,j1,i2;
820  
821 if ($COMP(con)!=4 && $COMP(con)!=8)
822 $CROAK("In ccNcompt, connectivity must be 4 or 8, you gave %d",$COMP(con));
823 loop(n,m) %{ /* Copy */
824 $b() = $a();
825 %}
826  
827 /* 1st pass counts max possible compts, 2nd records equivalences */
828  
829 for (pass = 0; pass<2; pass++) {
830  
831 if (pass==1) {
832 equiv = (PDL_Long*) malloc((newlabel+1)*sizeof(PDL_Long));
833 if (equiv==(PDL_Long*)0)
834 $CROAK("Out of memory");
835 for(i=0;i<=newlabel;i++)
836 equiv[i]=i;
837 }
838  
839 newlabel = 1; /* Running label */
840  
841 loop (n,m) %{ /* Loop over image pixels */
842 nfound = 0; /* Number of neighbour >0 */
843  
844 i1 = m-1; j1 = n-1; i2 = m+1; /*West x, North y, East x*/
845  
846 if ($b() > 0) { /* Check 4 neighbour already seen */
847  
848 if (m>0 && $b(m=>i1)>0) /*West*/
849 neighbour[nfound++] = $b(m=>i1); /* Store label of it */
850 if (n>0 && $b(n=>j1)>0) /*North*/
851 neighbour[nfound++] = $b(n=>j1);
852 if (n>0 && m>0 && $b(m=>i1, n=>j1)>0 && $COMP(con) == 8) /*North-West*/
853 neighbour[nfound++] = $b(m=>i1, n=>j1);
854 if (n>0 && m<($SIZE(m)-1) && $b(m=>i2, n=>j1)>0 && $COMP(con) == 8) /*North-East*/
855 neighbour[nfound++] = $b(m=>i2, n=>j1);
856  
857 if (nfound==0) { /* Assign new label */
858 $b() = newlabel++;
859 }
860 else {
861 $b() = neighbour[0];
862 if (nfound>1 && pass == 1) { /* Assign equivalents */
863 for(k=1; k<nfound; k++)
864 AddEquiv( equiv, (PDL_Long)$b(),
865 neighbour[k] );
866 }
867 }
868 }
869  
870 else { /* No label */
871  
872 $b() = 0;
873 }
874  
875 %} /* End of image loop */
876  
877 } /* Passes */
878  
879 /* Replace each cycle by single label */
880  
881 count = 0;
882 for (i = 1; i <= newlabel; i++)
883 if ( i <= equiv[i] ) {
884 count++;
885 this = i;
886 while ( equiv[this] != i ) {
887 next = equiv[this];
888 equiv[this] = count;
889 this = next;
890 }
891 equiv[this] = count;
892 }
893  
894  
895 /* Now remove equivalences */
896  
897 loop (n,m) %{ /* Loop over image pixels */
898 $b() = equiv[ (PDL_Long) $b() ] ;
899 %}
900  
901 free(equiv); /* Tidy */
902 ');
903  
904 pp_add_exported('polyfill');
905 pp_addpm(<<'EOPM');
906 =head2 polyfill
907  
908 =for ref
909  
910 fill the area of the given polygon with the given colour.
911  
912 This function works inplace, i.e. modifies C<im>.
913  
914 =for usage
915  
916 polyfill($im,$ps,$colour,[\%options]);
917  
918 The default method of determining which points lie inside of the polygon used
919 is not as strict as the method used in L</pnpoly>. Often, it includes vertices
920 and edge points. Set the C<Method> option to change this behaviour.
921  
922 =for option
923  
924 Method - Set the method used to determine which points lie in the polygon.
925 => Default - internal PDL algorithm
926 => pnpoly - use the L</pnpoly> algorithm
927  
928 =for example
929  
930 # Make a convex 3x3 square of 1s in an image using the pnpoly algorithm
931 $ps = pdl([3,3],[3,6],[6,6],[6,3]);
932 polyfill($im,$ps,1,{'Method' =>'pnpoly'});
933  
934 =cut
935 sub PDL::polyfill {
936 my $opt;
937 $opt = pop @_ if ref($_[-1]) eq 'HASH';
938 barf('Usage: polyfill($im,$ps,$colour,[\%options])') unless @_==3;
939 my ($im,$ps,$colour) = @_;
940  
941 if($opt) {
942 use PDL::Options qw();
943 my $parsed = PDL::Options->new({'Method' => undef});
944 $parsed->options($opt);
945 if( $parsed->current->{'Method'} eq 'pnpoly' ) {
946 PDL::pnpolyfill_pp($im,$ps,$colour);
947 }
948 }
949 else
950 {
951 PDL::polyfill_pp($im,$ps,$colour);
952 }
953 return $im;
954  
955 }
956  
957 *polyfill = \&PDL::polyfill;
958  
959 EOPM
960  
961 pp_add_exported('', 'pnpoly');
962 pp_addpm(<<'EOPM');
963  
964 =head2 pnpoly
965  
966 =for ref
967  
968 'points in a polygon' selection from a 2-D ndarray
969  
970 =for usage
971  
972 $mask = $img->pnpoly($ps);
973  
974 # Old style, do not use
975 $mask = pnpoly($x, $y, $px, $py);
976  
977 For a closed polygon determined by the sequence of points in {$px,$py}
978 the output of pnpoly is a mask corresponding to whether or not each
979 coordinate (x,y) in the set of test points, {$x,$y}, is in the interior
980 of the polygon. This is the 'points in a polygon' algorithm from
981 L<http://www.ecse.rpi.edu/Homepages/wrf/Research/Short_Notes/pnpoly.html>
982 and vectorized for PDL by Karl Glazebrook.
983  
984 =for example
985  
986 # define a 3-sided polygon (a triangle)
987 $ps = pdl([3, 3], [20, 20], [34, 3]);
988  
989 # $tri is 0 everywhere except for points in polygon interior
990 $tri = $img->pnpoly($ps);
991  
992 With the second form, the x and y coordinates must also be specified.
993 B< I<THIS IS MAINTAINED FOR BACKWARD COMPATIBILITY ONLY> >.
994  
995 $px = pdl( 3, 20, 34 );
996 $py = pdl( 3, 20, 3 );
997 $x = $img->xvals; # get x pixel coords
998 $y = $img->yvals; # get y pixel coords
999  
1000 # $tri is 0 everywhere except for points in polygon interior
1001 $tri = pnpoly($x,$y,$px,$py);
1002  
1003 =cut
1004  
1005 # From: http://www.ecse.rpi.edu/Homepages/wrf/Research/Short_Notes/pnpoly.html
1006 #
1007 # Fixes needed to pnpoly code:
1008 #
1009 # Use topdl() to ensure ndarray args
1010 #
1011 # Add POD docs for usage
1012 #
1013 # Calculate first term in & expression to use to fix divide-by-zero when
1014 # the test point is in line with a vertical edge of the polygon.
1015 # By adding the value of $mask we prevent a divide-by-zero since the &
1016 # operator does not "short circuit".
1017  
1018 sub PDL::pnpoly {
1019 barf('Usage: $mask = pnpoly($img, $ps);') unless(@_==2 || @_==4);
1020 my ($tx, $ty, $vertx, $verty) = @_;
1021  
1022 # if only two inputs, use the pure PP version
1023 unless( defined $vertx ) {
1024 barf("ps must contain pairwise points.\n") unless $ty->getdim(0) == 2;
1025  
1026 # Input mapping: $img => $tx, $ps => $ty
1027 return PDL::pnpoly_pp($tx,$ty);
1028 }
1029  
1030 my $testx = PDL::Core::topdl($tx)->dummy(0);
1031 my $testy = PDL::Core::topdl($ty)->dummy(0);
1032 my $vertxj = PDL::Core::topdl($vertx)->rotate(1);
1033 my $vertyj = PDL::Core::topdl($verty)->rotate(1);
1034 my $mask = ( ($verty>$testy) == ($vertyj>$testy) );
1035 my $c = sumover( ! $mask & ( $testx < ($vertxj-$vertx) * ($testy-$verty)
1036 / ($vertyj-$verty+$mask) + $vertx) ) % 2;
1037 return $c;
1038 }
1039  
1040 *pnpoly = \&PDL::pnpoly;
1041  
1042 EOPM
1043  
1044 pp_add_exported('', 'polyfillv');
1045 pp_addpm(<<'EOPM');
1046  
1047 =head2 polyfillv
1048  
1049 =for ref
1050  
1051 return the (dataflowed) area of an image described by a polygon
1052  
1053 =for usage
1054  
1055 polyfillv($im,$ps,[\%options]);
1056  
1057 The default method of determining which points lie inside of the polygon used
1058 is not as strict as the method used in L</pnpoly>. Often, it includes vertices
1059 and edge points. Set the C<Method> option to change this behaviour.
1060  
1061 =for option
1062  
1063 Method - Set the method used to determine which points lie in the polygon.
1064 => Default - internal PDL algorithm
1065 => pnpoly - use the L</pnpoly> algorithm
1066  
1067 =for example
1068  
1069 # increment intensity in area bounded by $poly using the pnpoly algorithm
1070 $im->polyfillv($poly,{'Method'=>'pnpoly'})++; # legal in perl >= 5.6
1071  
1072 # compute average intensity within area bounded by $poly using the default algorithm
1073 $av = $im->polyfillv($poly)->avg;
1074  
1075 =cut
1076  
1077 sub PDL::polyfillv :lvalue {
1078 my $opt;
1079 $opt = pop @_ if ref($_[-1]) eq 'HASH';
1080 barf('Usage: polyfillv($im,$ps,[\%options])') unless @_==2;
1081  
1082 my ($im,$ps) = @_;
1083 barf("ps must contain pairwise points.\n") unless $ps->getdim(0) == 2;
1084  
1085 if($opt) {
1086 use PDL::Options qw();
1087 my $parsed = PDL::Options->new({'Method' => undef});
1088 $parsed->options($opt);
1089 return $im->where(PDL::pnpoly_pp($im, $ps)) if $parsed->current->{'Method'} eq 'pnpoly';
1090 }
1091  
1092 PDL::polyfill_pp(my $msk = zeroes(long,$im->dims), $ps, 1);
1093 return $im->where($msk);
1094 }
1095 *polyfillv = \&PDL::polyfillv;
1096  
1097 EOPM
1098  
1099 pp_addhdr(<<'EOF');
1100 int getnewsize(PDL_Indx cols, PDL_Indx rows, float fangle, PDL_Indx *newcols, PDL_Indx *newrows);
1101 typedef unsigned char imT; /* image type */
1102 int rotate(imT *im, imT *out, int cols, int rows, int nc, int nr,
1103 float fangle, imT bgval, int antialias);
1104 EOF
1105 pp_add_exported('','rotnewsz');
1106 pp_addxs('
1107  
1108 void
1109 rotnewsz(m,n,angle)
1110 int m
1111 int n
1112 float angle
1113 PPCODE:
1114 PDL_Indx newcols, newrows;
1115  
1116 if (getnewsize(m,n,angle,&newcols,&newrows) != 0)
1117 croak("wrong angle (should be between -90 and +90)");
1118 EXTEND(sp,2);
1119 PUSHs(sv_2mortal(newSVnv(newcols)));
1120 PUSHs(sv_2mortal(newSVnv(newrows)));
1121 ');
1122  
1123 pp_def('rot2d',
1124 HandleBad => 0,
1125 Pars => 'im(m,n); float angle(); bg(); int aa(); [o] om(p,q)',
1126 GenericTypes => $A,
1127 Code => 'int ierr;
1128 if ((ierr = rotate($P(im),$P(om),$SIZE(m),$SIZE(n),$SIZE(p),
1129 $SIZE(q),$angle(),$bg(),$aa())) != 0) {
1130 if (ierr == -1)
1131 $CROAK("error during rotate, wrong angle");
1132 else
1133 $CROAK("wrong output dims, did you set them?");
1134 }',
1135 RedoDimsCode => '
1136 if (getnewsize($SIZE(m),$SIZE(n),
1137 $angle(), &$SIZE(p),
1138 &$SIZE(q)) != 0)
1139 $CROAK("error during rotate, wrong angle");
1140 /* printf("o: %d, p: %d\n",ncols,nrows); */
1141 ',
1142 GenericTypes => ['B'],
1143 Doc => << 'EOD',
1144  
1145 =for ref
1146  
1147 rotate an image by given C<angle>
1148  
1149 =for example
1150  
1151 # rotate by 10.5 degrees with antialiasing, set missing values to 7
1152 $rot = $im->rot2d(10.5,7,1);
1153  
1154 This function rotates an image through an C<angle> between -90 and + 90
1155 degrees. Uses/doesn't use antialiasing depending on the C<aa> flag.
1156 Pixels outside the rotated image are set to C<bg>.
1157  
1158 Code modified from pnmrotate (Copyright Jef Poskanzer) with an algorithm based
1159 on "A Fast Algorithm for General Raster Rotation" by Alan Paeth,
1160 Graphics Interface '86, pp. 77-81.
1161  
1162 Use the C<rotnewsz> function to find out about the dimension of the
1163 newly created image
1164  
1165 ($newcols,$newrows) = rotnewsz $oldn, $oldm, $angle;
1166  
1167 L<PDL::Transform> offers a more general interface to
1168 distortions, including rotation, with various types of sampling; but
1169 rot2d is faster.
1170  
1171 =cut
1172  
1173 EOD
1174 );
1175  
1176 pp_def('bilin2d',
1177 HandleBad => 0,
1178 Pars => 'Int(n,m); [io] O(q,p)',
1179 Doc=><<'EOD',
1180 =for ref
1181  
1182 Bilinearly maps the first ndarray in the second. The
1183 interpolated values are actually added to the second
1184 ndarray which is supposed to be larger than the first one.
1185 EOD
1186 GenericTypes => $A,
1187 RedoDimsCode =>'
1188 if ($SIZE(q)<$SIZE(n) || $SIZE(p)<$SIZE(m))
1189 $CROAK("the second matrix must be greater than first!");
1190 ',
1191 Code =>'
1192 PDL_CLDouble dx = ((PDL_CLDouble) ($SIZE(n)-1)) / ($SIZE(q)-1);
1193 PDL_CLDouble dy = ((PDL_CLDouble) ($SIZE(m)-1)) / ($SIZE(p)-1);
1194 broadcastloop %{
1195 PDL_CLDouble x=0;
1196 loop (q) %{
1197 PDL_CLDouble y = 0;
1198 loop (p) %{
1199 PDL_Indx ii = floor(x);
1200 if (ii>=($SIZE(n)-1)) ii = $SIZE(n)-2;
1201 PDL_Indx jj = floor(y);
1202 if (jj>=($SIZE(m)-1)) jj = $SIZE(m)-2;
1203 PDL_Indx ii1 = ii+1;
1204 PDL_Indx jj1 = jj+1;
1205 PDL_CLDouble y1 = $Int(n=>ii,m=>jj);
1206 PDL_CLDouble y2 = $Int(n=>ii1,m=>jj);
1207 PDL_CLDouble y3 = $Int(n=>ii1,m=>jj1);
1208 PDL_CLDouble y4 = $Int(n=>ii,m=>jj1);
1209 PDL_CLDouble t = x-ii;
1210 PDL_CLDouble u = y-jj;
1211 $O() += (1-t)*(1-u)*y1 + t*(1-u)*y2 + t*u*y3 + (1-t)*u*y4;
1212 y+=dy;
1213 %}
1214 x+=dx;
1215 %}
1216 %}
1217 ');
1218  
1219 pp_def('rescale2d',
1220 HandleBad => 0,
1221 Pars => 'Int(m,n); [io] O(p,q)',
1222 Doc=><<'EOD',
1223 =for ref
1224  
1225 The first ndarray is rescaled to the dimensions of the second
1226 (expanding or meaning values as needed) and then added to it in place.
1227 Nothing useful is returned.
1228  
1229 If you want photometric accuracy or automatic FITS header metadata
1230 tracking, consider using L<PDL::Transform::map|PDL::Transform/map>
1231 instead: it does these things, at some speed penalty compared to
1232 rescale2d.
1233 EOD
1234 GenericTypes => $A,
1235 Code =>'
1236 if($SIZE(p) >= $SIZE(m) && $SIZE(q) >= $SIZE(n)) {
1237 PDL_LDouble kx = ((PDL_LDouble) $SIZE(p)) / $SIZE(m);
1238 PDL_LDouble ky = ((PDL_LDouble) $SIZE(q)) / $SIZE(n);
1239 broadcastloop %{
1240 PDL_Indx lx = 0;
1241 loop (m) %{
1242 PDL_Indx ly = 0, cx = rint((m+1)*kx);
1243 loop (n) %{
1244 PDL_Indx cy = rint((n+1)*ky);
1245 loop (p=lx:cx,q=ly:cy) %{
1246 $O() += $Int();
1247 %}
1248 ly = cy + 1;
1249 %}
1250 lx = cx + 1;
1251 %}
1252 %}
1253 }
1254 else if($SIZE(p) < $SIZE(m) && $SIZE(q) < $SIZE(n)) {
1255 PDL_LDouble kx = ((PDL_LDouble) $SIZE(m)) / $SIZE(p);
1256 PDL_LDouble ky = ((PDL_LDouble) $SIZE(n)) / $SIZE(q);
1257 broadcastloop %{
1258 PDL_Indx lx = 0;
1259 loop (p) %{
1260 PDL_Indx ly = 0, cx = rint((p+1)*kx);
1261 loop (q) %{
1262 PDL_Indx cy = rint((q+1)*ky);
1263 PDL_LDouble temp = 0.0;
1264 PDL_Indx num = 0;
1265 loop (m=lx:cx,n=ly:cy) %{
1266 temp += $Int();
1267 num++;
1268 %}
1269 $O() += temp/num;
1270 ly = cy + 1;
1271 %}
1272 lx = cx + 1;
1273 %}
1274 %}
1275 }
1276 else if($SIZE(p) >= $SIZE(m) && $SIZE(q) < $SIZE(n)) {
1277 PDL_LDouble kx = ((PDL_LDouble) $SIZE(p)) / $SIZE(m);
1278 PDL_LDouble ky = ((PDL_LDouble) $SIZE(n)) / $SIZE(q);
1279 broadcastloop %{
1280 PDL_Indx lx = 0;
1281 loop (m) %{
1282 PDL_Indx ly = 0, cx = rint((m+1)*kx);
1283 loop (q) %{
1284 PDL_Indx cy = rint((q+1)*ky);
1285 PDL_LDouble temp = 0.0;
1286 PDL_Indx num = 0;
1287 loop (n=ly:cy) %{
1288 temp += $Int();
1289 num++;
1290 %}
1291 loop (p=lx:cx) %{
1292 $O() += temp/num;
1293 %}
1294 ly = cy + 1;
1295 %}
1296 lx = cx + 1;
1297 %}
1298 %}
1299 }
1300 else if($SIZE(p) < $SIZE(m) && $SIZE(q) >= $SIZE(n)) {
1301 PDL_LDouble kx = ((PDL_LDouble) $SIZE(m)) / $SIZE(p);
1302 PDL_LDouble ky = ((PDL_LDouble) $SIZE(q)) / $SIZE(n);
1303 broadcastloop %{
1304 PDL_Indx lx = 0;
1305 loop (p) %{
1306 PDL_Indx ly = 0, cx = rint((p+1)*kx);
1307 loop (n) %{
1308 PDL_Indx cy = rint((n+1)*ky);
1309 PDL_CLDouble temp = 0.0;
1310 PDL_Indx num = 0;
1311 loop (m=lx:cx) %{
1312 temp += $Int();
1313 num++;
1314 %}
1315 loop (q=ly:cy) %{
1316 $O() += temp/num;
1317 %}
1318 ly = cy + 1;
1319 %}
1320 lx = cx + 1;
1321 %}
1322 %}
1323 }
1324 else $CROAK("I am not supposed to be here, please report the bug to <chri@infis.univ.ts.it>");
1325 ');
1326  
1327 # functions to make handling 2D polynomial mappings a bit easier
1328 #
1329 pp_add_exported('', 'fitwarp2d applywarp2d');
1330  
1331 pp_addpm( '
1332  
1333 =head2 fitwarp2d
1334  
1335 =for ref
1336  
1337 Find the best-fit 2D polynomial to describe
1338 a coordinate transformation.
1339  
1340 =for usage
1341  
1342 ( $px, $py ) = fitwarp2d( $x, $y, $u, $v, $nf, { options } )
1343  
1344 Given a set of points in the output plane (C<$u,$v>), find
1345 the best-fit (using singular-value decomposition) 2D polynomial
1346 to describe the mapping back to the image plane (C<$x,$y>).
1347 The order of the fit is controlled by the C<$nf> parameter
1348 (the maximum power of the polynomial is C<$nf - 1>), and you
1349 can restrict the terms to fit using the C<FIT> option.
1350  
1351 C<$px> and C<$py> are C<np> by C<np> element ndarrays which describe
1352 a polynomial mapping (of order C<np-1>)
1353 from the I<output> C<(u,v)> image to the I<input> C<(x,y)> image:
1354  
1355 x = sum(j=0,np-1) sum(i=0,np-1) px(i,j) * u^i * v^j
1356 y = sum(j=0,np-1) sum(i=0,np-1) py(i,j) * u^i * v^j
1357  
1358 The transformation is returned for the reverse direction (ie
1359 output to input image) since that is what is required by the
1360 L</warp2d> routine. The L</applywarp2d>
1361 routine can be used to convert a set of C<$u,$v> points given
1362 C<$px> and C<$py>.
1363  
1364 Options:
1365  
1366 =for options
1367  
1368 FIT - which terms to fit? default ones(byte,$nf,$nf)
1369  
1370 =begin comment
1371  
1372 old option, caused trouble
1373 THRESH - in svd, remove terms smaller than THRESH * max value
1374 default is 1.0e-5
1375  
1376 =end comment
1377  
1378 =over 4
1379  
1380 =item FIT
1381  
1382 C<FIT> allows you to restrict which terms of the polynomial to fit:
1383 only those terms for which the FIT ndarray evaluates to true will be
1384 evaluated. If a 2D ndarray is sent in, then it is
1385 used for the x and y polynomials; otherwise
1386 C<< $fit->slice(":,:,(0)") >> will be used for C<$px> and
1387 C<< $fit->slice(":,:,(1)") >> will be used for C<$py>.
1388  
1389 =begin comment
1390  
1391 =item THRESH
1392  
1393 Remove all singular values whose value is less than C<THRESH>
1394 times the largest singular value.
1395  
1396 =end comment
1397  
1398 =back
1399  
1400 The number of points must be at least equal to the number of
1401 terms to fit (C<$nf*$nf> points for the default value of C<FIT>).
1402  
1403 =for example
1404  
1405 # points in original image
1406 $x = pdl( 0, 0, 100, 100 );
1407 $y = pdl( 0, 100, 100, 0 );
1408 # get warped to these positions
1409 $u = pdl( 10, 10, 90, 90 );
1410 $v = pdl( 10, 90, 90, 10 );
1411 #
1412 # shift of origin + scale x/y axis only
1413 $fit = byte( [ [1,1], [0,0] ], [ [1,0], [1,0] ] );
1414 ( $px, $py ) = fitwarp2d( $x, $y, $u, $v, 2, { FIT => $fit } );
1415 print "px = ${px}py = $py";
1416 px =
1417 [
1418 [-12.5 1.25]
1419 [ 0 0]
1420 ]
1421 py =
1422 [
1423 [-12.5 0]
1424 [ 1.25 0]
1425 ]
1426 #
1427 # Compared to allowing all 4 terms
1428 ( $px, $py ) = fitwarp2d( $x, $y, $u, $v, 2 );
1429 print "px = ${px}py = $py";
1430 px =
1431 [
1432 [ -12.5 1.25]
1433 [ 1.110223e-16 -1.1275703e-17]
1434 ]
1435 py =
1436 [
1437 [ -12.5 1.6653345e-16]
1438 [ 1.25 -5.8546917e-18]
1439 ]
1440  
1441 # A higher-degree polynomial should not affect the answer much, but
1442 # will require more control points
1443  
1444 $x = $x->glue(0,pdl(50,12.5, 37.5, 12.5, 37.5));
1445 $y = $y->glue(0,pdl(50,12.5, 37.5, 37.5, 12.5));
1446 $u = $u->glue(0,pdl(73,20,40,20,40));
1447 $v = $v->glue(0,pdl(29,20,40,40,20));
1448 ( $px3, $py3 ) = fitwarp2d( $x, $y, $u, $v, 3 );
1449 print "px3 =${px3}py3 =$py3";
1450 px3 =
1451 [
1452 [-6.4981162e+08 71034917 -726498.95]
1453 [ 49902244 -5415096.7 55945.388]
1454 [ -807778.46 88457.191 -902.51612]
1455 ]
1456 py3 =
1457 [
1458 [-6.2732159e+08 68576392 -701354.77]
1459 [ 48175125 -5227679.8 54009.114]
1460 [ -779821.18 85395.681 -871.27997]
1461 ]
1462  
1463 #This illustrates an important point about singular value
1464 #decompositions that are used in fitwarp2d: like all SVDs, the
1465 #rotation matrices are not unique, and so the $px and $py returned
1466 #by fitwarp2d are not guaranteed to be the "simplest" solution.
1467 #They do still work, though:
1468  
1469 ($x3,$y3) = applywarp2d($px3,$py3,$u,$v);
1470 print approx $x3,$x,1e-4;
1471 [1 1 1 1 1 1 1 1 1]
1472 print approx $y3,$y;
1473 [1 1 1 1 1 1 1 1 1]
1474  
1475 =head2 applywarp2d
1476  
1477 =for ref
1478  
1479 Transform a set of points using a 2-D polynomial mapping
1480  
1481 =for usage
1482  
1483 ( $x, $y ) = applywarp2d( $px, $py, $u, $v )
1484  
1485 Convert a set of points (stored in 1D ndarrays C<$u,$v>)
1486 to C<$x,$y> using the 2-D polynomial with coefficients stored in C<$px>
1487 and C<$py>. See L</fitwarp2d>
1488 for more information on the format of C<$px> and C<$py>.
1489  
1490 =cut
1491  
1492 # use SVD to fit data. Assuming no errors.
1493  
1494 =pod
1495  
1496 =begin comment
1497  
1498 Some explanation of the following three subroutines (_svd, _mkbasis,
1499 and fitwarp2d): See Wolberg 1990 (full ref elsewhere in this
1500 documentation), Chapter 3.6 "Polynomial Transformations". The idea is
1501 that, given a set of control points in the input and output images
1502 denoted by coordinates (x,y) and (u,v), we want to create a polynomial
1503 transformation that relates u to linear combinations of powers of x
1504 and y, and another that relates v to powers of x and y.
1505  
1506 The transformations used here and by Wolberg differ slightly, but the
1507 basic idea is something like this. For each u and each v, define a
1508 transform:
1509  
1510 u = (sum over j) (sum over i) a_{ij} x**i * y**j , (eqn 1)
1511 v = (sum over j) (sum over i) b_{ij} x**i * y**j . (eqn 2)
1512  
1513 We can write this in matrix notation. Given that there are M control
1514 points, U is a column vector with M rows. A and B are vectors containing
1515 the N coefficients (related to the degree of the polynomial fit). W
1516 is an MxN matrix of the basis row-vectors (the x**i y**j).
1517  
1518 The matrix equations we are trying to solve are
1519 U = W A , (eqn 3)
1520 V = W B . (eqn 4)
1521  
1522 We need to find the A and B column vectors, those are the coefficients
1523 of the polynomial terms in W. W is not square, so it has no inverse.
1524 But is has a pseudo-inverse W^+ that is NxM. We are going to use that
1525 pseudo-inverse to isolate A and B, like so:
1526  
1527 W^+ U = W^+ W A = A (eqn 5)
1528 W^+ V = W^+ W B = B (eqn 6)
1529  
1530 We are going to get W^+ by performing a singular value decomposition
1531 of W (to use some of the variable names below):
1532  
1533 W = $svd_u x SIGMA x $svd_v->transpose (eqn 7)
1534 W^+ = $svd_v x SIGMA^+ x $svd_u->transpose . (eqn 8)
1535  
1536 Here SIGMA is a square diagonal matrix that contains the singular
1537 values of W that are in the variable $svd_w. SIGMA^+ is the
1538 pseudo-inverse of SIGMA, which is calculated by replacing the non-zero
1539 singular values with their reciprocals, and then taking the transpose
1540 of the matrix (which is a no-op since the matrix is square and
1541 diagonal).
1542  
1543 So the code below does this:
1544  
1545 _mkbasis computes the matrix W, given control coordinates u and v and
1546 the maximum degree of the polynomial (and the terms to use).
1547  
1548 _svd takes the svd of W, computes the pseudo-inverse of W, and then
1549 multiplies that with the U vector (there called $y). The output of
1550 _svd is the A or B vector in eqns 5 & 6 above. Rarely is the matrix
1551 multiplication explicit, unfortunately, so I have added EXPLANATIONs
1552 below.
1553  
1554 =end comment
1555  
1556 =cut
1557  
1558 sub _svd ($$) {
1559 my $basis = shift;
1560 my $y = shift;
1561 # my $thresh = shift;
1562  
1563 # if we had errors for these points, would normalise the
1564 # basis functions, and the output array, by these errors here
1565  
1566 # perform the SVD
1567 my ( $svd_u, $svd_w, $svd_v ) = svd( $basis );
1568  
1569 # DAL, 09/2017: the reason for removing ANY singular values, much less
1570 #the smallest ones, is not clear. I commented the line below since this
1571 #actually removes the LARGEST values in SIGMA^+.
1572 # $svd_w *= ( $svd_w >= ($svd_w->max * $thresh ) );
1573 # The line below would instead remove the SMALLEST values in SIGMA^+, but I can see no reason to include it either.
1574 # $svd_w *= ( $svd_w <= ($svd_w->min / $thresh ) );
1575  
1576 # perform the back substitution
1577 # EXPLANATION: same thing as $svd_u->transpose x $y->transpose.
1578 my $tmp = $y x $svd_u;
1579  
1580 #EXPLANATION: the division by (the non-zero elements of) $svd_w (the singular values) is
1581 #equivalent to $sigma_plus x $tmp, so $tmp is now SIGMA^+ x $svd_u x $y
1582 $tmp /= $svd_w->setvaltobad(0.0);
1583 $tmp->inplace->setbadtoval(0.0);
1584  
1585 #EXPLANATION: $svd_v x SIGMA^+ x $svd_u x $y
1586 return sumover( $svd_v * $tmp );
1587  
1588 } # sub: _svd()
1589  
1590 #_mkbasis returns an ndarray in which the k(=j*n+i)_th column is v**j * u**i
1591 #k=0 j=0 i=0
1592 #k=1 j=0 i=1
1593 #k=2 j=0 i=2
1594 #k=3 j=1 i=0
1595 # ...
1596  
1597 #each row corresponds to a control point
1598 #and the rows for each of the control points look like this, e.g.:
1599 # (1) (u) (u**2) (v) (vu) (v(u**2)) (v**2) ((v**2)u) ((v**2)(u**2))
1600 # and so on for the next control point.
1601  
1602 sub _mkbasis ($$$$) {
1603 my $fit = shift; # dims n n
1604 my $npts = shift; # scalar
1605 my $u = shift; # dims npts
1606 my $v = shift; # dims npts
1607 my $ncoeff = sum( $fit );
1608 my $fit_coords = $fit->whichND; # dims uv ncoeff
1609 cat($u,$v) # npts uv
1610 ->transpose # uv npts
1611 ->dummy(1,$ncoeff) # uv ncoeff npts
1612 ->ipow($fit_coords) # same
1613 ->prodover # ncoeff npts
1614 ;
1615 } # sub: _mkbasis()
1616  
1617 sub PDL::fitwarp2d {
1618 croak "Usage: (\$px,\$py) = fitwarp2d(x(m);y(m);u(m);v(m);\$nf; { options })"
1619 if $#_ < 4 or ( $#_ >= 5 and ref($_[5]) ne "HASH" );
1620  
1621 my $x = shift;
1622 my $y = shift;
1623 my $u = shift;
1624 my $v = shift;
1625 my $nf = shift;
1626  
1627 my $opts = PDL::Options->new( { FIT => ones(byte,$nf,$nf) } ); #, THRESH => 1.0e-5 } );
1628 $opts->options( $_[0] ) if $#_ > -1;
1629 my $oref = $opts->current();
1630  
1631 # safety checks
1632 my $npts = $x->nelem;
1633 croak "fitwarp2d: x, y, u, and v must be the same size (and 1D)"
1634 unless $npts == $y->nelem and $npts == $u->nelem and $npts == $v->nelem
1635 and $x->getndims == 1 and $y->getndims == 1 and $u->getndims == 1 and $v->getndims == 1;
1636  
1637 # my $svd_thresh = $$oref{THRESH};
1638 # croak "fitwarp2d: THRESH option must be >= 0."
1639 # if $svd_thresh < 0;
1640  
1641 my $fit = $$oref{FIT};
1642 my $fit_ndim = $fit->getndims();
1643 croak "fitwarp2d: FIT option must be sent a (\$nf,\$nf[,2]) element ndarray"
1644 unless UNIVERSAL::isa($fit,"PDL") and
1645 ($fit_ndim == 2 or ($fit_ndim == 3 and $fit->getdim(2) == 2)) and
1646 $fit->getdim(0) == $nf and $fit->getdim(1) == $nf;
1647  
1648 # how many coeffs to fit (first we ensure $fit is either 0 or 1)
1649 $fit = convert( $fit != 0, byte );
1650  
1651 my ( $fitx, $fity, $ncoeffx, $ncoeffy, $ncoeff );
1652 if ( $fit_ndim == 2 ) {
1653 $fitx = $fit;
1654 $fity = $fit;
1655 $ncoeff = $ncoeffx = $ncoeffy = sum( $fit );
1656 } else {
1657 $fitx = $fit->slice(",,(0)");
1658 $fity = $fit->slice(",,(1)");
1659 $ncoeffx = sum($fitx);
1660 $ncoeffy = sum($fity);
1661 $ncoeff = $ncoeffx > $ncoeffy ? $ncoeffx : $ncoeffy;
1662 }
1663  
1664 croak "fitwarp2d: number of points ($npts) must be >= \$ncoeff ($ncoeff)"
1665 unless $npts >= $ncoeff;
1666  
1667 # create the basis functions for the SVD fitting
1668 my ( $basisx, $basisy );
1669  
1670 $basisx = _mkbasis( $fitx, $npts, $u, $v );
1671  
1672 if ( $fit_ndim == 2 ) {
1673 $basisy = $basisx;
1674 } else {
1675 $basisy = _mkbasis( $fity, $npts, $u, $v );
1676 }
1677  
1678 my $px = _svd( $basisx, $x ); # $svd_thresh);
1679 my $py = _svd( $basisy, $y ); # $svd_thresh);
1680  
1681 # convert into $nf x $nf element ndarrays, if necessary
1682 my $nf2 = $nf * $nf;
1683  
1684 return ( $px->reshape($nf,$nf), $py->reshape($nf,$nf) )
1685 if $ncoeff == $nf2 and $ncoeffx == $ncoeffy;
1686  
1687 # re-create the matrix
1688 my $xtmp = zeroes( $nf, $nf );
1689 my $ytmp = zeroes( $nf, $nf );
1690  
1691 my $kx = 0;
1692 my $ky = 0;
1693 foreach my $i ( 0 .. ($nf - 1) ) {
1694 foreach my $j ( 0 .. ($nf - 1) ) {
1695 if ( $fitx->at($i,$j) ) {
1696 $xtmp->set($i,$j, $px->at($kx) );
1697 $kx++;
1698 }
1699 if ( $fity->at($i,$j) ) {
1700 $ytmp->set($i,$j, $py->at($ky) );
1701 $ky++;
1702 }
1703 }
1704 }
1705  
1706 return ( $xtmp, $ytmp )
1707  
1708 } # sub: fitwarp2d
1709  
1710 *fitwarp2d = \&PDL::fitwarp2d;
1711  
1712 sub PDL::applywarp2d {
1713 # checks
1714 croak "Usage: (\$x,\$y) = applywarp2d(px(nf,nf);py(nf,nf);u(m);v(m);)"
1715 if $#_ != 3;
1716  
1717 my $px = shift;
1718 my $py = shift;
1719 my $u = shift;
1720 my $v = shift;
1721 my $npts = $u->nelem;
1722  
1723 # safety check
1724 croak "applywarp2d: u and v must be the same size (and 1D)"
1725 unless $npts == $u->nelem and $npts == $v->nelem
1726 and $u->getndims == 1 and $v->getndims == 1;
1727  
1728 my $nf = $px->getdim(0);
1729 my $nf2 = $nf * $nf;
1730  
1731 # could remove terms with 0 coeff here
1732 # (would also have to remove them from px/py for
1733 # the matrix multiplication below)
1734 #
1735 my $mat = _mkbasis( ones(byte,$nf,$nf), $npts, $u, $v );
1736  
1737 my $x = reshape( $mat x $px->flat->transpose(), $npts );
1738 my $y = reshape( $mat x $py->flat->transpose(), $npts );
1739 return ( $x, $y );
1740  
1741 } # sub: applywarp2d
1742  
1743 *applywarp2d = \&PDL::applywarp2d;
1744  
1745 ' );
1746  
1747 ## resampling routines taken from v3.6-0 of the Eclipse package
1748 ## http://www.eso.org/eclipse by Nicolas Devillard
1749 ##
1750  
1751 pp_addhdr( '#include "resample.h"' . "\n" );
1752  
1753 pp_def(
1754 'warp2d',
1755 Doc=> <<'EOD',
1756 =for ref
1757  
1758 Warp a 2D image given a polynomial describing the I<reverse> mapping.
1759  
1760 =for usage
1761  
1762 $out = warp2d( $img, $px, $py, { options } );
1763  
1764 Apply the polynomial transformation encoded in the C<$px> and
1765 C<$py> ndarrays to warp the input image C<$img> into the output
1766 image C<$out>.
1767  
1768 The format for the polynomial transformation is described in
1769 the documentation for the L</fitwarp2d> routine.
1770  
1771 At each point C<x,y>, the closest 16 pixel values are combined
1772 with an interpolation kernel to calculate the value at C<u,v>.
1773 The interpolation is therefore done in the image, rather than
1774 Fourier, domain.
1775 By default, a C<tanh> kernel is used, but this can be changed
1776 using the C<KERNEL> option discussed below
1777 (the choice of kernel depends on the frequency content of the input image).
1778  
1779 The routine is based on the C<warping> command from
1780 the Eclipse data-reduction package - see http://www.eso.org/eclipse/ - and
1781 for further details on image resampling see
1782 Wolberg, G., "Digital Image Warping", 1990, IEEE Computer
1783 Society Press ISBN 0-8186-8944-7).
1784  
1785 Currently the output image is the same size as the input one,
1786 which means data will be lost if the transformation reduces
1787 the pixel scale. This will (hopefully) be changed soon.
1788  
1789 =for example
1790  
1791 $img = rvals(byte,501,501);
1792 imag $img, { JUSTIFY => 1 };
1793 #
1794 # use a not-particularly-obvious transformation:
1795 # x = -10 + 0.5 * $u - 0.1 * $v
1796 # y = -20 + $v - 0.002 * $u * $v
1797 #
1798 $px = pdl( [ -10, 0.5 ], [ -0.1, 0 ] );
1799 $py = pdl( [ -20, 0 ], [ 1, 0.002 ] );
1800 $wrp = warp2d( $img, $px, $py );
1801 #
1802 # see the warped image
1803 imag $warp, { JUSTIFY => 1 };
1804  
1805 The options are:
1806  
1807 =for options
1808  
1809 KERNEL - default value is tanh
1810 NOVAL - default value is 0
1811  
1812 C<KERNEL> is used to specify which interpolation kernel to use
1813 (to see what these kernels look like, use the
1814 L</warp2d_kernel> routine).
1815 The options are:
1816  
1817 =over 4
1818  
1819 =item tanh
1820  
1821 Hyperbolic tangent: the approximation of an ideal box filter by the
1822 product of symmetric tanh functions.
1823  
1824 =item sinc
1825  
1826 For a correctly sampled signal, the ideal filter in the fourier domain is a rectangle,
1827 which produces a C<sinc> interpolation kernel in the spatial domain:
1828  
1829 sinc(x) = sin(pi * x) / (pi * x)
1830  
1831 However, it is not ideal for the C<4x4> pixel region used here.
1832  
1833 =item sinc2
1834  
1835 This is the square of the sinc function.
1836  
1837 =item lanczos
1838  
1839 Although defined differently to the C<tanh> kernel, the result is very
1840 similar in the spatial domain. The Lanczos function is defined as
1841  
1842 L(x) = sinc(x) * sinc(x/2) if abs(x) < 2
1843 = 0 otherwise
1844  
1845 =item hann
1846  
1847 This kernel is derived from the following function:
1848  
1849 H(x) = a + (1-a) * cos(2*pi*x/(N-1)) if abs(x) < 0.5*(N-1)
1850 = 0 otherwise
1851  
1852 with C<a = 0.5> and N currently equal to 2001.
1853  
1854 =item hamming
1855  
1856 This kernel uses the same C<H(x)> as the Hann filter, but with
1857 C<a = 0.54>.
1858  
1859 =back
1860  
1861 C<NOVAL> gives the value used to indicate that a pixel in the
1862 output image does not map onto one in the input image.
1863 EOD
1864 HandleBad => 0,
1865 Pars => 'img(m,n); ldouble px(np,np); ldouble py(np,np); [o] warp(m,n); ldouble [t] poly(np); ldouble [t] kernel(ns)',
1866 OtherPars => 'char *kernel_type; double noval; IV nsamples => ns',
1867 GenericTypes => [ qw(F D E) ],
1868 PMCode => <<'EOF',
1869 # support routine
1870 {
1871 my %warp2d = map { ($_,1) } qw( tanh sinc sinc2 lanczos hamming hann );
1872 # note: convert to lower case
1873 sub _check_kernel ($$) {
1874 my $kernel = lc shift;
1875 my $code = shift;
1876 barf "Unknown kernel $kernel sent to $code\n" .
1877 "\tmust be one of [" . join(',',keys %warp2d) . "]\n"
1878 unless exists $warp2d{$kernel};
1879 return $kernel;
1880 }
1881 }
1882  
1883 sub PDL::warp2d {
1884 my $opts = PDL::Options->new( { KERNEL => "tanh", NOVAL => 0 } );
1885 $opts->options( pop(@_) ) if ref($_[$#_]) eq "HASH";
1886 die "Usage: warp2d( in(m,n), px(np,np); py(np,np); [o] out(m,n), {Options} )"
1887 if $#_<2 || $#_>3;
1888 my $img = shift;
1889 my $px = shift;
1890 my $py = shift;
1891 my $out = $#_ == -1 ? PDL->null() : shift;
1892 # safety checks
1893 my $copt = $opts->current();
1894 my $kernel = _check_kernel( $$copt{KERNEL}, "warp2d" );
1895 &PDL::_warp2d_int( $img, $px, $py, $out, $kernel, $$copt{NOVAL}, _get_kernel_size() );
1896 return $out;
1897 }
1898 EOF
1899 Code => <<'EOF',
1900 PDL_Indx k, lx_3, ly_3, px, py, tabx, taby;
1901 PDL_Indx da[16], db[16] ;
1902 PDL_LDouble cur, neighbors[16], rsc[8], sumrs, x, y, *kernel = $P(kernel), *poly = $P(poly);
1903  
1904 /* Generate interpolation kernel */
1905 if (!generate_interpolation_kernel($COMP(kernel_type), $SIZE(ns), kernel))
1906 $CROAK("Invalid kernel type '%s'", $COMP(kernel_type));
1907  
1908 /* Compute sizes */
1909 lx_3 = $SIZE(m) - 3;
1910 ly_3 = $SIZE(n) - 3;
1911  
1912 /* Pre compute leaps for 16 closest neighbors positions */
1913 da[0] = -1; db[0] = -1;
1914 da[1] = 0; db[1] = -1;
1915 da[2] = 1; db[2] = -1;
1916 da[3] = 2; db[3] = -1;
1917  
1918 da[4] = -1; db[4] = 0;
1919 da[5] = 0; db[5] = 0;
1920 da[6] = 1; db[6] = 0;
1921 da[7] = 2; db[7] = 0;
1922  
1923 da[8] = -1; db[8] = 1;
1924 da[9] = 0; db[9] = 1;
1925 da[10] = 1; db[10] = 1;
1926 da[11] = 2; db[11] = 1;
1927  
1928 da[12] = -1; db[12] = 2;
1929 da[13] = 0; db[13] = 2;
1930 da[14] = 1; db[14] = 2;
1931 da[15] = 2; db[15] = 2;
1932  
1933 poly[0] = 1.0;
1934  
1935 /* Loop over the output image */
1936 broadcastloop %{
1937 loop(n) %{
1938  
1939 /* fill in poly array */
1940 loop (np=1) %{
1941 poly[np] = (PDL_LDouble) n * poly[np-1];
1942 %}
1943  
1944 loop(m) %{
1945  
1946 /* Compute the original source for this pixel */
1947 x = poly2d_compute( $SIZE(np), $P(px), m, poly );
1948 y = poly2d_compute( $SIZE(np), $P(py), m, poly );
1949  
1950 /* Which is the closest integer positioned neighbor? */
1951 px = (int)x ;
1952 py = (int)y ;
1953  
1954 if ((px < 1) || (px > lx_3) || (py < 1) || (py > ly_3))
1955 $warp() = ($GENERIC()) $COMP(noval);
1956 else {
1957  
1958 /* Now feed the positions for the closest 16 neighbors */
1959 for (k=0 ; k<16 ; k++) {
1960 neighbors[k] = (PDL_LDouble) $img( m => px+da[k], n => py+db[k] );
1961 }
1962  
1963 /* Which tabulated value index shall we use? */
1964 tabx = (x - (PDL_LDouble)px) * (PDL_LDouble)(TABSPERPIX) ;
1965 taby = (y - (PDL_LDouble)py) * (PDL_LDouble)(TABSPERPIX) ;
1966  
1967 /* Compute resampling coefficients */
1968 /* rsc[0..3] in x, rsc[4..7] in y */
1969  
1970 rsc[0] = kernel[TABSPERPIX + tabx] ;
1971 rsc[1] = kernel[tabx] ;
1972 rsc[2] = kernel[TABSPERPIX - tabx] ;
1973 rsc[3] = kernel[2 * TABSPERPIX - tabx] ;
1974 rsc[4] = kernel[TABSPERPIX + taby] ;
1975 rsc[5] = kernel[taby] ;
1976 rsc[6] = kernel[TABSPERPIX - taby] ;
1977 rsc[7] = kernel[2 * TABSPERPIX - taby] ;
1978  
1979 sumrs = (rsc[0]+rsc[1]+rsc[2]+rsc[3]) *
1980 (rsc[4]+rsc[5]+rsc[6]+rsc[7]) ;
1981  
1982 /* Compute interpolated pixel now */
1983 cur = rsc[4] * ( rsc[0]*neighbors[0] +
1984 rsc[1]*neighbors[1] +
1985 rsc[2]*neighbors[2] +
1986 rsc[3]*neighbors[3] ) +
1987 rsc[5] * ( rsc[0]*neighbors[4] +
1988 rsc[1]*neighbors[5] +
1989 rsc[2]*neighbors[6] +
1990 rsc[3]*neighbors[7] ) +
1991 rsc[6] * ( rsc[0]*neighbors[8] +
1992 rsc[1]*neighbors[9] +
1993 rsc[2]*neighbors[10] +
1994 rsc[3]*neighbors[11] ) +
1995 rsc[7] * ( rsc[0]*neighbors[12] +
1996 rsc[1]*neighbors[13] +
1997 rsc[2]*neighbors[14] +
1998 rsc[3]*neighbors[15] ) ;
1999  
2000 /* Copy the value to the output image */
2001 $warp() = ($GENERIC()) (cur/sumrs);
2002  
2003 } /* if: edge or interior */
2004  
2005 %} /* loop(m) */
2006 %} /* loop(n) */
2007 %} /* broadcastloop */
2008 EOF
2009 ); # pp_def: warp2d
2010  
2011 pp_addxs( '
2012  
2013 int
2014 _get_kernel_size()
2015 PROTOTYPE:
2016 CODE:
2017 RETVAL = KERNEL_SAMPLES;
2018 OUTPUT:
2019 RETVAL
2020  
2021 ');
2022  
2023 pp_def( 'warp2d_kernel',
2024 Doc => <<'EOF',
2025 =for ref
2026  
2027 Return the specified kernel, as used by L</warp2d>
2028  
2029 =for usage
2030  
2031 ( $x, $k ) = warp2d_kernel( $name )
2032  
2033 The valid values for C<$name> are the same as the C<KERNEL> option
2034 of L</warp2d>.
2035  
2036 =for example
2037  
2038 line warp2d_kernel( "hamming" );
2039 EOF
2040 HandleBad => 0,
2041 PMCode => '
2042 sub PDL::warp2d_kernel ($) {
2043 my $kernel = _check_kernel( shift, "warp2d_kernel" );
2044 &PDL::_warp2d_kernel_int( my $x=PDL->null, my $k=PDL->null, $kernel, _get_kernel_size() );
2045 return ( $x, $k );
2046 }
2047 ',
2048 Pars => '[o] x(n); [o] k(n); ldouble [t] kernel(n)',
2049 OtherPars => 'char *name; PDL_Indx nsize => n',
2050 GenericTypes => [ qw(F D E) ],
2051 Code => <<'EOF',
2052 PDL_LDouble *kernel = $P(kernel);
2053 if (!generate_interpolation_kernel($COMP(name), $SIZE(n), kernel))
2054 $CROAK("Invalid kernel type '%s'", $COMP(name));
2055 /* fill in ndarrays */
2056 PDL_LDouble xx = 0.0;
2057 broadcastloop %{
2058 loop (n) %{
2059 $x() = xx;
2060 $k() = kernel[n];
2061 xx += 1.0 / (double) TABSPERPIX;
2062 %}
2063 %}
2064 EOF
2065 );
2066  
2067 pp_done();
« lib/PDL/ImageND-pp-convolveND.c ↑ Summary lib/PDL/MatrixOps-pp-squaretotri.c »