void imProcessCanny(const imImage* im, imImage* NewImage, float stddev)
{
int width = 1;
float **smx,**smy;
float **dx,**dy;
int i;
float gau[MAX_MASK_SIZE], dgau[MAX_MASK_SIZE];
/* Create a Gaussian and a derivative of Gaussian filter mask */
for(i=0; i<MAX_MASK_SIZE; i++)
{
gau[i] = meanGauss ((float)i, stddev);
if (gau[i] < 0.005)
{
width = i;
break;
}
dgau[i] = dGauss ((float)i, stddev);
}
smx = f2d (im->height, im->width);
smy = f2d (im->height, im->width);
/* Convolution of source image with a Gaussian in X and Y directions */
seperable_convolution (im, gau, width, smx, smy);
MAG_SCALE = 0;
/* Now convolve smoothed data with a derivative */
dx = f2d (im->height, im->width);
dxy_seperable_convolution (smx, im->height, im->width, dgau, width, dx, 1);
free(smx[0]); free(smx);
dy = f2d (im->height, im->width);
dxy_seperable_convolution (smy, im->height, im->width, dgau, width, dy, 0);
free(smy[0]); free(smy);
if (MAG_SCALE)
MAG_SCALE = 255.0f/(1.4142f*MAG_SCALE);
/* Non-maximum suppression - edge pixels should be a local max */
nonmax_suppress (dx, dy, NewImage);
free(dx[0]); free(dx);
free(dy[0]); free(dy);
}
SEXP declust(SEXP theta,
SEXP rbwd,
SEXP revents,
SEXP rpoly,
SEXP tperiod)
{
SEXP dim, pdim, out, integ0;
// extract events
PROTECT(dim = allocVector(INTSXP, 2));
dim = getAttrib(revents, R_DimSymbol);
int N = INTEGER(dim)[0];
double *events = REAL(revents);
double t[N], x[N], y[N], m[N], bk[N], pb[N], lam[N];
for (int i = 0; i < N; i++)
{
t[i] = events[i];
x[i] = events[N + i];
y[i] = events[2 * N + i];
m[i] = events[3 * N + i];
bk[i] = events[5 * N + i];
pb[i] = events[6 * N + i];
lam[i] = events[7 * N + i];
}
// extract polygon information
PROTECT(pdim = allocVector(INTSXP, 2));
pdim = getAttrib(rpoly, R_DimSymbol);
int np = INTEGER(pdim)[0];
double *poly = REAL(rpoly);
double px[np], py[np];
for (int i = 0; i < np; i++)
{
px[i] = poly[i];
py[i] = poly[np + i];
}
// extract time period information
double *tper = REAL(tperiod);
double tstart2 = tper[0], tlength = tper[1];
// extract bandwidthes
double *bwd = REAL(rbwd);
// extract model paramters
double *tht = REAL(theta);
double s, r0, w[1];
for (int i = 0; i < N; i++)
{
s = 0;
for (int j = 0; j < N; j++)
{
r0 = dist(x[i], y[i], x[j], y[j]);
s += pb[j] * dGauss(r0, bwd[j]);
}
bk[i] = s / (tlength - tstart2);
events[5 * N + i] = bk[i];
}
s = 0;
for (int i = 0; i < N; i++)
{
w[0] = bwd[i];
s += pb[i] * polyinteg(pGauss, w, &np, px, py, x[i], y[i]);
lam[i] = lambdaj(tht,i, t, x, y, m, bk);
events[6 * N + i] = (tht[0] * tht[0] * bk[i]) / lam[i];
events[7 * N + i] = lam[i];
}
PROTECT(out = allocVector(VECSXP, 2));
PROTECT(integ0 = allocVector(REALSXP, 1));
double *integ0P = REAL(integ0);
integ0P[0] = s;
SET_VECTOR_ELT(out, 0, revents);
SET_VECTOR_ELT(out, 1, integ0);
UNPROTECT(4);
return(out);
}
bool IPLCanny::processInputData(IPLImage* image , int, bool useOpenCV)
{
// delete previous result
delete _result;
_result = NULL;
delete _binaryImage;
_binaryImage = NULL;
int width = image->width();
int height = image->height();
_result = new IPLImage( image->type(), width, height );
_binaryImage = new IPLImage( IPLData::IMAGE_BW, width, height );
// get properties
int window = getProcessPropertyInt("window");
double sigma = getProcessPropertyDouble("sigma");
double lowThreshold = getProcessPropertyDouble("lowThreshold");
double highThreshold = getProcessPropertyDouble("highThreshold");
std::stringstream s;
s << "Window: ";
s << window;
addInformation(s.str());
//! @todo currently only the opencv implementation works
if(useOpenCV || true)
{
notifyProgressEventHandler(-1);
cv::Mat input;
cv::Mat output;
cvtColor(image->toCvMat(), input, CV_BGR2GRAY);
cv::Canny(input, output, lowThreshold*255, highThreshold*255, window);
delete _result;
_result = new IPLImage(output);
return true;
}
return false;
// Create a Gaussian 1D filter
int N = ceil( sigma * sqrt( 2.0*log( 1.0/0.015 ) ) + 1.0 );
double ssq = sigma*sigma;
double* gau = new double [window];
double* dgau = new double [window];
for( int k = -N; k <= N; ++k )
{
gau[k+N] = gauss ( (double)k, ssq );
dgau[k+N] = dGauss ( (double)k, 0, ssq );
}
// Create a directional derivative of 2D Gaussian (along X-axis)
// Since the result is symmetric along X, we can get the derivative along
// Y-axis simply by transposing the result for X direction.
// DoubleImage* dgau = new DoubleImage( window, window );
// for( int y = -N; y <= N; ++y )
// for( int x = -N; x <= N; ++x )
// dgau->f(x+N, y+N) = dGauss( x, y, ssq );
int progress = 0;
int maxProgress = width * image->getNumberOfPlanes();
int nrOfPlanes = image->getNumberOfPlanes();
//#pragma omp parallel for
for( int planeNr=0; planeNr < nrOfPlanes; planeNr++ )
{
IPLImagePlane* plane = image->plane( planeNr );
IPLImagePlane* newplane = _result->plane( planeNr );
// ******** Gaussian filtering of input image
IPLImagePlane* gI = new IPLImagePlane( width, height );
// horizontal run (normalizing original image)
IPLImagePlane* tmpI = new IPLImagePlane( width, height );
for(int x=0; x<width; x++)
{ // progress
notifyProgressEventHandler(100*progress++/maxProgress);
for(int y=0; y<height; y++)
{
double sum = 0;
int i = 0;
for( int kx=-N; kx<=N; kx++ )
{
double img = (double) plane->bp(x+kx, y);
sum += (img * gau[i++]);
}
tmpI->p(x,y) = (double) (sum);
}
}
// vertiacl run
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
double sum = 0;
int i = 0;
for( int ky=-N; ky<=N; ky++ )
{
double img = tmpI->bp(x, y+ky);
sum += (img * gau[i++]);
}
gI->p(x,y) = sum;
}
}
//delete tmpI;
// ******** Apply directional derivatives ...
// ... in x-direction
IPLImagePlane* dx = new IPLImagePlane( width, height );
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
dx->p(x,y) = 0.0;
for( int k=1; k<N; k++ )
{
dx->p(x,y) += ( gI->bp(x-k,y) - gI->bp(x+k,y) ) * dgau[k];
}
}
}
// double maxVal = 0.0;
// for(int x=0; x<width; x++)
// for(int y=0; y<height; y++)
// if( dx->f(x,y) > maxVal ) maxVal = dx->f(x,y);
// ... in y-direction
IPLImagePlane* dy = new IPLImagePlane( width, height );
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
dy->p(x,y) = 0.0;
for( int k=1; k<N; k++ )
{
dy->p(x,y) += ( gI->bp(x,y-k) - gI->bp(x,y+k) ) * dgau[k];
}
}
}
// ******** Compute magnitude and binarization thresholds
IPLImagePlane* mag = new IPLImagePlane( width, height );
double magMax = 0.0;
double magMin = 999999999.0;
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
double val = sqrt( dx->p(x,y)*dx->p(x,y) + dy->p(x,y)*dy->p(x,y) );
mag->p(x,y) = val;
if( val > magMax ) magMax = val;
if( val < magMin ) magMin = val;
}
}
//// ******** Non-maxima suppression - edge pixels should be a local maximum
_orientedImage = new IPLOrientedImage( width, height );
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
double ix = dx->p(x,y);
double iy = dy->p(x,y);
double g = mag->p(x,y);
// determine 4-neighbor direction of gradient
int dir4 = 0;
if( (iy<=0.0 && ix>-iy) || (iy>=0.0 && ix<-iy) )
dir4 = 1;
else if( (ix>0.0 && -iy>=ix) || (ix<0.0 && -iy<=ix) )
dir4 = 2;
else if( (ix<=0.0 && ix>iy) || (ix>=0.0 && ix<iy) )
dir4 = 3;
else if( (iy<0.0 && ix<=iy) || (iy>0.0 && ix>=iy) )
dir4 = 4;
else
continue;
double gradmag1, gradmag2, d;
switch(dir4)
{
case 1: d = std::fabs(iy/ix);
gradmag1 = mag->bp(x+1,y)*(1-d) + mag->bp(x+1,y-1)*d;
gradmag2 = mag->bp(x-1,y)*(1-d) + mag->bp(x-1,y+1)*d;
break;
case 2: d = std::fabs(ix/iy);
gradmag1 = mag->bp(x,y-1)*(1-d) + mag->bp(x+1,y-1)*d;
gradmag2 = mag->bp(x,y+1)*(1-d) + mag->bp(x-1,y+1)*d;
break;
case 3: d = std::fabs(ix/iy);
gradmag1 = mag->bp(x,y-1)*(1-d) + mag->bp(x-1,y-1)*d;
gradmag2 = mag->bp(x,y+1)*(1-d) + mag->bp(x+1,y+1)*d;
break;
case 4: d = std::fabs(iy/ix);
gradmag1 = mag->bp(x-1,y)*(1-d) + mag->bp(x-1,y-1)*d;
gradmag2 = mag->bp(x+1,y)*(1-d) + mag->bp(x+1,y+1)*d;
break;
}
if( g > gradmag1 && g > gradmag2 )
{
_orientedImage->magnitude(x,y) = g;
_orientedImage->phase(x,y) = atan2(iy,ix);
}
}
}
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
_orientedImage->magnitude(x,y) /= magMax;
double val = _orientedImage->magnitude(x,y)*255.0;
// double val = mag->f(x,y)/magMax*255.0;
if (val > 255.0 ) val = 255.0;
if (val < 0.0 ) val = 0.0;
newplane->p(x,y) = (unsigned char ) val;
}
}
// ******** Binarize with hysteresis threshold
double hist[ 256 ];
for( int i=0; i<256; ++i )
hist[i] = 0;
int pixCount = 0;
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
if( _orientedImage->magnitude(x,y) > 0.0 )
{
int index = floor( _orientedImage->magnitude(x,y)*256.0+0.5 );
++hist[ index ];
++pixCount;
}
}
}
double PercentOfPixelsNotEdges = 0.7*pixCount;
double highThresh = 0.0;
double cumsum = 0.0;
for( int i=0; i<256; ++i )
{
cumsum += hist[i];
if( cumsum > PercentOfPixelsNotEdges )
{
highThresh = (double)i / 256.0;
break;
}
}
double lowThresh = 0.4 * highThresh;
IPLImagePlane* binPlane = _binaryImage->plane( 0 );
for(int x=0; x<width; x++)
{
for(int y=0; y<height; y++)
{
if(_orientedImage->magnitude(x,y) >= highThresh)
trace(x, y, lowThresh, _orientedImage, binPlane);
}
}
//delete dx;
//delete dy;
//delete gI;
thinning(_orientedImage, binPlane, newplane );
}
//delete [] gau;
//delete [] dgau;
return true;
}
