public double[] calcResponse(int channel, int lambda) throws TypeNotSupportedException {
int N = settings.getNoOfImages();
int width = imp.getWidth();
int height = imp.getHeight();
int type = imp.getType();
double[] out = new double[width * height];
// number of saturated pixels
int saturated_pixels = 0;
// number of levels
int M = settings.getLevels(); // settings.getZmax() - settings.getZmin() + 1;
// response function
double[] I[] = new double[3][M];
// initialize response function linear
I[channel] = ResponseFunctionTools.responseLinear(M);
// previous response function
double[] Ip = new double[M];
// indizes
int m, i, j;
// normalize
I[channel] = normalizeI(I[channel], M);
// previous response (copied for first convergence check)
for (m = 0; m < M; m++) {
Ip[m] = I[channel][m];
}
// long start = System.currentTimeMillis();
RobertsonRadiance r =
new RobertsonRadiance(
imp, new ResponseFunction(I, imp.getType(), w), settings.getExpTimes());
out = r.calcRadiance(channel);
// System.out.println("first response end: " + (System.currentTimeMillis() - start));
// Optimization process
boolean converged = false;
long cardEm[] = new long[M];
double[] sum = new double[M];
int cur_it = 0;
double pdelta = 0.0;
// start = System.currentTimeMillis();
while (!converged) {
// 1. Minimize with respect to I
for (m = 0; m < M; m++) {
cardEm[m] = 0;
sum[m] = 0.0;
}
for (i = 0; i < N; i++) {
Object pixels = imp.getImageStack().getPixels(i + 1);
double ti = settings.getExpTimes()[i];
for (j = 0; j < width * height; j++) {
m = ImageJTools.getPixelValue(pixels, j, type, channel);
if (m < M && m >= 0) {
sum[m] += ti * out[j];
cardEm[m]++;
} else {
System.out.println("robertson02: m out of range: " + m);
}
}
}
for (m = 0; m < M; m++) {
if (cardEm[m] != 0) {
I[channel][m] = sum[m] / cardEm[m];
} else {
I[channel][m] = 0.0;
}
}
// System.out.println("minimize end: " + (System.currentTimeMillis() - start));
// start = System.currentTimeMillis();
// 2. Normalize I
I[channel] = normalizeI(I[channel], M);
// System.out.println("normalize end: " + (System.currentTimeMillis() - start));
// start = System.currentTimeMillis();
// 3. Apply new response
r.setR(new ResponseFunction(I, imp.getType(), w));
out = r.calcRadiance(channel);
// System.out.println("apply response end: " + (System.currentTimeMillis() - start));
// start = System.currentTimeMillis();
// 4. Check stopping condition
double delta = 0.0;
int hits = 0;
for (m = 0; m < M; m++) {
if (I[channel][m] != 0.0) {
double diff = I[channel][m] - Ip[m];
delta += diff * diff;
Ip[m] = I[channel][m];
hits++;
}
}
delta /= hits;
// System.out.println("check stopping end: " + (System.currentTimeMillis() - start));
System.out.println(
" # " + cur_it + " delta = " + delta + " coverage: " + 100 * hits / M + "%");
if (delta < MAX_DELTA) {
converged = true;
} else if (Double.isNaN(delta) || (cur_it > MAXIT && pdelta < delta)) {
System.out.println("algorithm failed to converge, too noisy data in range\n");
break;
}
pdelta = delta;
cur_it++;
}
for (int u = 0; u < I[channel].length; u++) {
System.out.println(I[channel][u]);
}
return I[channel];
}
public RobertsonCalculator(ImagePlus imp, ResponseFunctionCalculatorSettings settings) {
this.imp = imp;
this.settings = settings;
this.w = new GaussianWeightFunction(settings.getZmin(), settings.getZmax(), 8.0);
}