Files
2026-01-11 10:16:12 +01:00

349 lines
12 KiB
C

/*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% %
% %
% %
% GGGG EEEEE M M %
% G E MM MM %
% G GG EEE M M M %
% G G E M M %
% GGGG EEEEE M M %
% %
% %
% Graphic Gems - Graphic Support Methods %
% %
% Software Design %
% Cristy %
% August 1996 %
% %
% %
% Copyright @ 1999 ImageMagick Studio LLC, a non-profit organization %
% dedicated to making software imaging solutions freely available. %
% %
% You may not use this file except in compliance with the License. You may %
% obtain a copy of the License at %
% %
% https://imagemagick.org/license/ %
% %
% Unless required by applicable law or agreed to in writing, software %
% distributed under the License is distributed on an "AS IS" BASIS, %
% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. %
% See the License for the specific language governing permissions and %
% limitations under the License. %
% %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%
%
*/
/*
Include declarations.
*/
#include "MagickCore/studio.h"
#include "MagickCore/color-private.h"
#include "MagickCore/draw.h"
#include "MagickCore/gem.h"
#include "MagickCore/gem-private.h"
#include "MagickCore/image.h"
#include "MagickCore/image-private.h"
#include "MagickCore/log.h"
#include "MagickCore/memory_.h"
#include "MagickCore/pixel-accessor.h"
#include "MagickCore/quantum.h"
#include "MagickCore/quantum-private.h"
#include "MagickCore/random_.h"
#include "MagickCore/resize.h"
#include "MagickCore/transform.h"
#include "MagickCore/signature-private.h"
/*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% %
% %
% %
% E x p a n d A f f i n e %
% %
% %
% %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% ExpandAffine() computes the affine's expansion factor, i.e. the square root
% of the factor by which the affine transform affects area. In an affine
% transform composed of scaling, rotation, shearing, and translation, returns
% the amount of scaling.
%
% The format of the ExpandAffine method is:
%
% double ExpandAffine(const AffineMatrix *affine)
%
% A description of each parameter follows:
%
% o expansion: ExpandAffine returns the affine's expansion factor.
%
% o affine: A pointer the affine transform of type AffineMatrix.
%
*/
MagickExport double ExpandAffine(const AffineMatrix *affine)
{
assert(affine != (const AffineMatrix *) NULL);
return(sqrt(fabs(affine->sx*affine->sy-affine->rx*affine->ry)));
}
/*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% %
% %
% %
% G e n e r a t e D i f f e r e n t i a l N o i s e %
% %
% %
% %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% GenerateDifferentialNoise() generates differential noise.
%
% The format of the GenerateDifferentialNoise method is:
%
% double GenerateDifferentialNoise(RandomInfo *random_info,
% const Quantum pixel,const NoiseType noise_type,const double attenuate)
%
% A description of each parameter follows:
%
% o random_info: the random info.
%
% o pixel: noise is relative to this pixel value.
%
% o noise_type: the type of noise.
%
% o attenuate: attenuate the noise.
%
*/
MagickPrivate double GenerateDifferentialNoise(RandomInfo *random_info,
const Quantum pixel,const NoiseType noise_type,const double attenuate)
{
#define SigmaUniform (attenuate*0.015625)
#define SigmaGaussian (attenuate*0.015625)
#define SigmaImpulse (attenuate*0.1)
#define SigmaLaplacian (attenuate*0.0390625)
#define SigmaMultiplicativeGaussian (attenuate*0.5)
#define SigmaPoisson (attenuate*12.5)
#define SigmaRandom (attenuate)
#define TauGaussian (attenuate*0.078125)
double
alpha,
beta,
noise,
sigma;
alpha=GetPseudoRandomValue(random_info);
switch (noise_type)
{
case UniformNoise:
default:
{
noise=(double) pixel+(double) QuantumRange*SigmaUniform*(alpha-0.5);
break;
}
case GaussianNoise:
{
double
gamma,
tau;
if (fabs(alpha) < MagickEpsilon)
alpha=1.0;
beta=GetPseudoRandomValue(random_info);
gamma=sqrt(-2.0*log(alpha));
sigma=gamma*cos((double) (2.0*MagickPI*beta));
tau=gamma*sin((double) (2.0*MagickPI*beta));
noise=(double) pixel+sqrt((double) pixel)*SigmaGaussian*sigma+
(double) QuantumRange*TauGaussian*tau;
break;
}
case ImpulseNoise:
{
if (alpha < (SigmaImpulse/2.0))
noise=0.0;
else
if (alpha >= (1.0-(SigmaImpulse/2.0)))
noise=(double) QuantumRange;
else
noise=(double) pixel;
break;
}
case LaplacianNoise:
{
if (alpha <= 0.5)
{
if (alpha <= MagickEpsilon)
noise=(double) (pixel-QuantumRange);
else
noise=(double) pixel+(double) QuantumRange*SigmaLaplacian*
log(2.0*alpha)+0.5;
break;
}
beta=1.0-alpha;
if (beta <= (0.5*MagickEpsilon))
noise=(double) (pixel+QuantumRange);
else
noise=(double) pixel-(double) QuantumRange*SigmaLaplacian*
log(2.0*beta)+0.5;
break;
}
case MultiplicativeGaussianNoise:
{
sigma=1.0;
if (alpha > MagickEpsilon)
sigma=sqrt(-2.0*log(alpha));
beta=GetPseudoRandomValue(random_info);
noise=(double) pixel+(double) pixel*SigmaMultiplicativeGaussian*sigma*
cos((double) (2.0*MagickPI*beta))/2.0;
break;
}
case PoissonNoise:
{
double
poisson;
ssize_t
i;
poisson=exp(-SigmaPoisson*QuantumScale*(double) pixel);
for (i=0; alpha > poisson; i++)
{
beta=GetPseudoRandomValue(random_info);
alpha*=beta;
}
noise=(double) QuantumRange*i*MagickSafeReciprocal(SigmaPoisson);
break;
}
case RandomNoise:
{
noise=(double) QuantumRange*SigmaRandom*alpha;
break;
}
}
return(noise);
}
/*
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% %
% %
% %
% G e t O p t i m a l K e r n e l W i d t h %
% %
% %
% %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% GetOptimalKernelWidth() computes the optimal kernel radius for a convolution
% filter. Start with the minimum value of 3 pixels and walk out until we drop
% below the threshold of one pixel numerical accuracy.
%
% The format of the GetOptimalKernelWidth method is:
%
% size_t GetOptimalKernelWidth(const double radius,
% const double sigma)
%
% A description of each parameter follows:
%
% o width: GetOptimalKernelWidth returns the optimal width of a
% convolution kernel.
%
% o radius: the radius of the Gaussian, in pixels, not counting the center
% pixel.
%
% o sigma: the standard deviation of the Gaussian, in pixels.
%
*/
MagickPrivate size_t GetOptimalKernelWidth1D(const double radius,
const double sigma)
{
double
alpha,
beta,
gamma,
normalize,
value;
size_t
width;
ssize_t
i,
j;
if (IsEventLogging() != MagickFalse)
(void) LogMagickEvent(TraceEvent,GetMagickModule(),"...");
if (radius > MagickEpsilon)
return((size_t) (2.0*ceil(radius)+1.0));
gamma=fabs(sigma);
if (gamma <= MagickEpsilon)
return(3UL);
alpha=MagickSafeReciprocal(2.0*gamma*gamma);
beta=(double) MagickSafeReciprocal((double) MagickSQ2PI*gamma);
for (width=5; ; )
{
normalize=0.0;
j=(ssize_t) (width-1)/2;
for (i=(-j); i <= j; i++)
normalize+=exp(-((double) (i*i))*alpha)*beta;
value=exp(-((double) (j*j))*alpha)*beta/normalize;
if ((value < QuantumScale) || (value < MagickEpsilon))
break;
width+=2;
}
return((size_t) (width-2));
}
MagickPrivate size_t GetOptimalKernelWidth2D(const double radius,
const double sigma)
{
double
alpha,
beta,
gamma,
normalize,
value;
size_t
width;
ssize_t
j,
u,
v;
if (IsEventLogging() != MagickFalse)
(void) LogMagickEvent(TraceEvent,GetMagickModule(),"...");
if (radius > MagickEpsilon)
return((size_t) (2.0*ceil(radius)+1.0));
gamma=fabs(sigma);
if (gamma <= MagickEpsilon)
return(3UL);
alpha=MagickSafeReciprocal(2.0*gamma*gamma);
beta=(double) MagickSafeReciprocal((double) Magick2PI*gamma*gamma);
for (width=5; ; )
{
normalize=0.0;
j=(ssize_t) (width-1)/2;
for (v=(-j); v <= j; v++)
for (u=(-j); u <= j; u++)
normalize+=exp(-((double) (u*u+v*v))*alpha)*beta;
value=exp(-((double) (j*j))*alpha)*beta/normalize;
if ((value < QuantumScale) || (value < MagickEpsilon))
break;
width+=2;
}
return((size_t) (width-2));
}
MagickPrivate size_t GetOptimalKernelWidth(const double radius,
const double sigma)
{
return(GetOptimalKernelWidth1D(radius,sigma));
}