/**
* Get standard deviation of the image
*
* @param img input image
* @return sigma
*/
protected double getStdDev(MTBImage img) {
int sizeStack = img.getSizeStack();
int sizeX = img.getSizeX();
int sizeY = img.getSizeY();
double val;
double mu = 0.0;
double mu2 = 0.0;
double N = 0.0;
double sigma;
for (int i = 0; i < sizeStack; i++) {
img.setCurrentSliceIndex(i);
for (int y = 0; y < sizeY; y++) {
for (int x = 0; x < sizeX; x++) {
val = img.getValueDouble(x, y);
mu += val;
mu2 += val * val;
N++;
}
}
}
mu /= N;
mu2 /= N;
sigma = Math.sqrt(mu2 - mu * mu);
return sigma;
}
/**
* Denoise wavelet coefficients using Jeffrey's noninformative prior for a given sigma of noise
*
* @param img input image
* @param sigma sigma of noise
*/
protected void denoise(MTBImage img, double sigma) {
int sizeStack = img.getSizeStack();
int sizeX = img.getSizeX();
int sizeY = img.getSizeY();
double s2 = 3.0 * sigma * sigma;
double val, val2;
for (int i = 0; i < sizeStack; i++) {
img.setCurrentSliceIndex(i);
for (int y = 0; y < sizeY; y++) {
for (int x = 0; x < sizeX; x++) {
val = img.getValueDouble(x, y);
val2 = (val * val - s2);
if (val2 < 0.0) val2 = 0.0;
if (val != 0.0) img.putValueDouble(x, y, val2 / val);
else img.putValueDouble(x, y, 0.0);
}
}
}
img.setCurrentSliceIndex(0);
}
/**
* Create an image with Gaussian noise
*
* @param mean
* @param sigma
* @param clippingFactor
* @param bins
* @param sizeX
* @param sizeY
* @param sizeZ
* @param sizeT
* @param sizeC
* @return
*/
protected MTBImage createGaussianNoiseImage(
double mean,
double sigma,
double clippingFactor,
int bins,
int sizeX,
int sizeY,
int sizeZ,
int sizeT,
int sizeC) {
MTBImage gImg =
MTBImage.createMTBImage(sizeX, sizeY, sizeZ, sizeT, sizeC, MTBImageType.MTB_DOUBLE);
int sizeStack = gImg.getSizeStack();
double[] dist = new double[bins];
double cs = -clippingFactor * sigma;
double gFactor = 1.0 / (Math.sqrt(2.0 * Math.PI) * sigma);
double lastVal = 0.0;
double X;
// cumulative distribution
for (int i = 0; i < bins; i++) {
X = ((double) i / (double) (bins - 1)) * 2.0 * cs - cs;
dist[i] = lastVal + gFactor * Math.exp(-0.5 * (X * X) / (sigma * sigma));
lastVal = dist[i];
}
// normalization
// for (int i = 0; i < bins; i++) {
// dist[i] /= dist[bins-1];
// }
double sample;
for (int i = 0; i < sizeStack; i++) {
gImg.setCurrentSliceIndex(i);
for (int y = 0; y < sizeY; y++) {
for (int x = 0; x < sizeX; x++) {
sample = getSample(dist);
sample = sample * 2.0 * cs - cs - mean;
gImg.putValueDouble(x, y, sample);
}
}
}
gImg.setCurrentSliceIndex(0);
return gImg;
}
