NAME
PDL::ImageND - useful image processing in N dimensions
DESCRIPTION
These routines act on PDLs as N-dimensional objects, not as broadcasted sets of 0-D or 1-D objects. The file is sort of a catch-all for broadly functional routines, most of which could legitimately be filed elsewhere (and probably will, one day).
ImageND is not a part of the PDL core (v2.4) and hence must be explicitly loaded.
SYNOPSIS
use PDL::ImageND;
$y = $x->convolveND($kernel,{bound=>'periodic'});
$y = $x->rebin(50,30,10);FUNCTIONS
convolve
Signature: (a(m); b(n); indx adims(p); indx bdims(q); [o]c(m))N-dimensional convolution (Deprecated; use convolveND)
$new = convolve $x, $kernelConvolve an array with a kernel, both of which are N-dimensional. This routine does direct convolution (by copying) but uses quasi-periodic boundary conditions: each dim "wraps around" to the next higher row in the next dim.
This routine is kept for backwards compatibility with earlier scripts; for most purposes you want convolveND instead: it runs faster and handles a variety of boundary conditions.
convolve does not process bad values. It will set the bad-value flag of all output ndarrays if the flag is set for any of the input ndarrays.
ninterpol()
N-dimensional interpolation routine
Signature: ninterpol(point(),data(n),[o]value()) $value = ninterpol($point, $data);ninterpol uses interpol to find a linearly interpolated value in N dimensions, assuming the data is spread on a uniform grid. To use an arbitrary grid distribution, need to find the grid-space point from the indexing scheme, then call ninterpol -- this is far from trivial (and ill-defined in general).
See also interpND, which includes boundary conditions and allows you to switch the method of interpolation, but which runs somewhat slower.
rebin
Signature: (a(m); [o]b(n); int ns => n)N-dimensional rebinning algorithm
$new = rebin $x, $dim1, $dim2,..;.
$new = rebin $x, $template;
$new = rebin $x, $template, {Norm => 1};Rebin an N-dimensional array to newly specified dimensions. Specifying `Norm' keeps the sum constant, otherwise the intensities are kept constant. If more template dimensions are given than for the input pdl, these dimensions are created; if less, the final dimensions are maintained as they were.
So if $x is a 10 x 10 pdl, then rebin($x,15) is a 15 x 10 pdl, while rebin($x,15,16,17) is a 15 x 16 x 17 pdl (where the values along the final dimension are all identical).
Expansion is performed by sampling; reduction is performed by averaging. If you want different behavior, use PDL::Transform::map instead. PDL::Transform::map runs slower but is more flexible.
rebin does not process bad values. It will set the bad-value flag of all output ndarrays if the flag is set for any of the input ndarrays.
circ_mean_p
Calculates the circular mean of an n-dim image and returns the projection. Optionally takes the center to be used.
$cmean=circ_mean_p($im);
$cmean=circ_mean_p($im,{Center => [10,10]});circ_mean
Smooths an image by applying circular mean. Optionally takes the center to be used.
circ_mean($im);
circ_mean($im,{Center => [10,10]});kernctr
`centre' a kernel (auxiliary routine to fftconvolve)
$kernel = kernctr($image,$smallk);
fftconvolve($image,$kernel);kernctr centres a small kernel to emulate the behaviour of the direct convolution routines.
convolveND
Signature: (k0(); pdl *k; pdl *aa; pdl *a)Speed-optimized convolution with selectable boundary conditions
$new = convolveND($x, $kernel, [ {options} ]);Convolve an array with a kernel, both of which are N-dimensional.
If the kernel has fewer dimensions than the array, then the extra array dimensions are broadcasted over. There are options that control the boundary conditions and method used.
The kernel's origin is taken to be at the kernel's center. If your kernel has a dimension of even order then the origin's coordinates get rounded up to the next higher pixel (e.g. (1,2) for a 3x4 kernel). This mimics the behavior of the earlier "convolve" and fftconvolve routines, so convolveND is a drop-in replacement for them.
The kernel may be any size compared to the image, in any dimension.
The kernel and the array are not quite interchangeable (as in mathematical convolution): the code is inplace-aware only for the array itself, and the only allowed boundary condition on the kernel is truncation.
convolveND is inplace-aware: say convolveND(inplace $x ,$k) to modify a variable in-place. You don't reduce the working memory that way -- only the final memory.
OPTIONS
Options are parsed by PDL::Options, so unique abbreviations are accepted.
- boundary (default: 'truncate')
-
The boundary condition on the array, which affects any pixel closer to the edge than the half-width of the kernel.
The boundary conditions are the same as those accepted by range, because this option is passed directly into range. Useful options are 'truncate' (the default), 'extend', and 'periodic'. You can select different boundary conditions for different axes -- see range for more detail.
The (default) truncate option marks all the near-boundary pixels as BAD if you have bad values compiled into your PDL and the array's badflag is set.
- method (default: 'auto')
-
The method to use for the convolution. Acceptable alternatives are 'direct', 'fft', or 'auto'. The direct method is an explicit copy-and-multiply operation; the fft method takes the Fourier transform of the input and output kernels. The two methods give the same answer to within double-precision numerical roundoff. The fft method is much faster for large kernels; the direct method is faster for tiny kernels. The tradeoff occurs when the array has about 400x more pixels than the kernel.
The default method is 'auto', which chooses direct or fft convolution based on the size of the input arrays.
NOTES
At the moment there's no way to broadcast over kernels. That could/should be fixed.
The broadcasting over input is cheesy and should probably be fixed: currently the kernel just gets dummy dimensions added to it to match the input dims. That does the right thing tersely but probably runs slower than a dedicated broadcastloop.
The direct copying code uses PP primarily for the generic typing: it includes its own broadcastloops.
convolveND does not process bad values. It will set the bad-value flag of all output ndarrays if the flag is set for any of the input ndarrays.
contour_segments
Signature: (c(); data(m,n); points(d,m,n);
[o] segs(d,q=CALC(($SIZE(m)-1)*($SIZE(n)-1)*4)); indx [o] cnt();)Finds a contour in given data. Takes 3 ndarrays as input:
$c is the contour value (broadcast with this)
$data is an [m,n] array of values at each point
$points is a list of [d,m,n] points. It should be a grid monotonically increasing with m and n.
Returns $segs, and $cnt which is the highest 2nd-dim index in $segs that's defined. The contours are a collection of disconnected line segments rather than a set of closed polygons.
The data array represents samples of some field observed on the surface described by points. This uses a variant of the Marching Squares algorithm, though without being data-driven.
contour_segments does not process bad values. It will set the bad-value flag of all output ndarrays if the flag is set for any of the input ndarrays.
contour_polylines
Signature: (c(); data(m,n); points(d,m,n);
indx [o] pathendindex(q=CALC(($SIZE(m)-1)*($SIZE(n)-1)*5)); [o] paths(d,q);
byte [t] seenmap(m,n))Finds polylines describing contours in given data. Takes 3 ndarrays as input:
$c is the contour value (broadcast with this)
$data is an [m,n] array of values at each point
$points is a list of [d,m,n] points. It should be a grid monotonically increasing with m and n.
Returns $pathendindex, and $paths. Any $pathendindex entries after the pointers to the ends of polylines are negative.
Algorithm
Has two modes: scanning, and line-walking. Scanning is done from the top left, along each row. Each point can be considered as, at a:
a|b
+-+-
c|d|eEvery potential boundary above, or to the left of (including the bottom boundaries), a has been cleared (marked with a space above).
Boundary detection
This is done by first checking both points' coordinates are within bounds, then checking if the boundary is marked seen, then detecting whether the two cells' values cross the contour threshold.
Scanning
If detect boundary between a-b, and also a-d, d-e, or b-e, line-walking starts a-b facing south.
If not, mark a-b seen.
If detect boundary a-d and c-d, line-walking starts a-d facing west.
If detect boundary a-d and also d-e or b-e, line-walking starts a-d facing east.
If not, mark a-d seen, and continue scanning.
Line-walking
The conditions above guarantee that any line started will have at least two points, since two connected "points" (boundaries between two cells) have been detected. The coordinates of the back end of the starting "point" (boundary with direction) are recorded.
At each, a line-point is emitted and that "point" is marked seen. The coordinates emitted are linearly interpolated between the coordinates of the two cells similarly to the Marching Squares algorithm.
The next "point" is sought, looking in order right, straight ahead, then left. Each one not detected is marked seen. That order means the walked boundary will always turn as much right (go clockwise) as available, thereby guaranteeing enclosing the area, which deals with saddle points.
If a next "point" is found, move to that and repeat.
If not, then if the front of the ending "point" (boundary plus direction) is identical to the back of the starting point, a final point is emitted to close the shape. Then the polyline is closed by emitting the current point-counter into polyendindex.
use PDL;
use PDL::ImageND;
use PDL::Graphics::Simple;
$SIZE = 500;
$vals = rvals($SIZE,$SIZE)->divide($SIZE/12.5)->sin;
@cntr_threshes = zeroes(9)->xlinvals($vals->minmax)->list;
$win = pgswin();
$xrange = [0,$vals->dim(0)-1]; $yrange = [0,$vals->dim(1)-1];
$win->plot(with=>'image', $vals, {xrange=>$xrange,yrange=>$yrange,j=>1},);
for $thresh (@cntr_threshes) {
($pi, $p) = contour_polylines($thresh, $vals, $vals->ndcoords);
$pi_max = $pi->max;
next if $pi_max < 0;
$pi = $pi->where($pi > -1);
$p = $p->slice(',0:'.$pi_max);
@paths = path_segs($pi, $p->mv(0,-1));
$win->oplot(
(map +(with=>'lines', $_->dog), @paths),
{xrange=>$xrange,yrange=>$yrange,j=>1},
);
}
print "ret> "; <>;contour_polylines does not process bad values. It will set the bad-value flag of all output ndarrays if the flag is set for any of the input ndarrays.
path_join
Signature: (e(v=2,n);
indx [o] pathendindex(n); indx [o] paths(nout=CALC($SIZE(n)*2));
indx [t] highestoutedge(d); indx [t] outedges(d,d); byte [t] hasinward(d);
indx [t] sourceids(d);
; PDL_Indx d => d; int directed)Links a (by default directed) graph's edges into paths.
The outputs are the indices into paths ending each path. Past the last path, the indices are set to -1.
path_join does not process bad values. It will set the bad-value flag of all output ndarrays if the flag is set for any of the input ndarrays.
path_segs
Divide a path into segments.
@segments = path_segs($pathindices, $paths);Returns a series of slices of the paths, such as those created by "path_join", stopping at the first negative index. Currently does not broadcast.
use PDL;
use PDL::ImageND;
use PDL::Graphics::Simple;
$SIZE = 500;
$vals = rvals($SIZE,$SIZE)->divide($SIZE/12.5)->sin;
@cntr_threshes = zeroes(9)->xlinvals($vals->minmax)->list;
$win = pgswin();
$xrange = [0,$vals->dim(0)-1]; $yrange = [0,$vals->dim(1)-1];
$win->plot(with=>'image', $vals, {xrange=>$xrange,yrange=>$yrange,j=>1},);
for $thresh (@cntr_threshes) {
my ($segs, $cnt) = contour_segments($thresh, $vals, $vals->ndcoords);
my $segscoords = $segs->slice(',0:'.$cnt->max)->clump(-1)->splitdim(0,4);
$linesegs = $segscoords->splitdim(0,2);
$uniqcoords = $linesegs->uniqvec;
next if $uniqcoords->dim(1) < 2;
$indexed = vsearchvec($linesegs, $uniqcoords)->uniqvec;
@paths = path_segs(path_join($indexed, $uniqcoords->dim(1), 0));
@paths = map $uniqcoords->dice_axis(1, $_)->mv(0,-1), @paths;
$win->oplot(
(map +(with=>'lines', $_->dog), @paths),
{xrange=>$xrange,yrange=>$yrange,j=>1},
);
}
print "ret> "; <>;AUTHORS
Copyright (C) Karl Glazebrook and Craig DeForest, 1997, 2003 All rights reserved. There is no warranty. You are allowed to redistribute this software / documentation under certain conditions. For details, see the file COPYING in the PDL distribution. If this file is separated from the PDL distribution, the copyright notice should be included in the file.