use of org.apache.commons.math3.fitting.PolynomialCurveFitter in project GDSC-SMLM by aherbert.
the class MeanVarianceTest method run.
/*
* (non-Javadoc)
*
* @see ij.plugin.PlugIn#run(java.lang.String)
*/
public void run(String arg) {
SMLMUsageTracker.recordPlugin(this.getClass(), arg);
if (Utils.isExtraOptions()) {
ImagePlus imp = WindowManager.getCurrentImage();
if (imp.getStackSize() > 1) {
GenericDialog gd = new GenericDialog(TITLE);
gd.addMessage("Perform single image analysis on the current image?");
gd.addNumericField("Bias", _bias, 0);
gd.showDialog();
if (gd.wasCanceled())
return;
singleImage = true;
_bias = Math.abs(gd.getNextNumber());
} else {
IJ.error(TITLE, "Single-image mode requires a stack");
return;
}
}
List<ImageSample> images;
String inputDirectory = "";
if (singleImage) {
IJ.showStatus("Loading images...");
images = getImages();
if (images.size() == 0) {
IJ.error(TITLE, "Not enough images for analysis");
return;
}
} else {
inputDirectory = IJ.getDirectory("Select image series ...");
if (inputDirectory == null)
return;
SeriesOpener series = new SeriesOpener(inputDirectory, false, 0);
series.setVariableSize(true);
if (series.getNumberOfImages() < 3) {
IJ.error(TITLE, "Not enough images in the selected directory");
return;
}
if (!IJ.showMessageWithCancel(TITLE, String.format("Analyse %d images, first image:\n%s", series.getNumberOfImages(), series.getImageList()[0]))) {
return;
}
IJ.showStatus("Loading images");
images = getImages(series);
if (images.size() < 3) {
IJ.error(TITLE, "Not enough images for analysis");
return;
}
if (images.get(0).exposure != 0) {
IJ.error(TITLE, "First image in series must have exposure 0 (Bias image)");
return;
}
}
boolean emMode = (arg != null && arg.contains("em"));
GenericDialog gd = new GenericDialog(TITLE);
gd.addMessage("Set the output options:");
gd.addCheckbox("Show_table", showTable);
gd.addCheckbox("Show_charts", showCharts);
if (emMode) {
// Ask the user for the camera gain ...
gd.addMessage("Estimating the EM-gain requires the camera gain without EM readout enabled");
gd.addNumericField("Camera_gain (ADU/e-)", cameraGain, 4);
}
gd.showDialog();
if (gd.wasCanceled())
return;
showTable = gd.getNextBoolean();
showCharts = gd.getNextBoolean();
if (emMode) {
cameraGain = gd.getNextNumber();
}
IJ.showStatus("Computing mean & variance");
final double nImages = images.size();
for (int i = 0; i < images.size(); i++) {
IJ.showStatus(String.format("Computing mean & variance %d/%d", i + 1, images.size()));
images.get(i).compute(singleImage, i / nImages, (i + 1) / nImages);
}
IJ.showProgress(1);
IJ.showStatus("Computing results");
// Allow user to input multiple bias images
int start = 0;
Statistics biasStats = new Statistics();
Statistics noiseStats = new Statistics();
final double bias;
if (singleImage) {
bias = _bias;
} else {
while (start < images.size()) {
ImageSample sample = images.get(start);
if (sample.exposure == 0) {
biasStats.add(sample.means);
for (PairSample pair : sample.samples) {
noiseStats.add(pair.variance);
}
start++;
} else
break;
}
bias = biasStats.getMean();
}
// Get the mean-variance data
int total = 0;
for (int i = start; i < images.size(); i++) total += images.get(i).samples.size();
if (showTable && total > 2000) {
gd = new GenericDialog(TITLE);
gd.addMessage("Table output requires " + total + " entries.\n \nYou may want to disable the table.");
gd.addCheckbox("Show_table", showTable);
gd.showDialog();
if (gd.wasCanceled())
return;
showTable = gd.getNextBoolean();
}
TextWindow results = (showTable) ? createResultsWindow() : null;
double[] mean = new double[total];
double[] variance = new double[mean.length];
Statistics gainStats = (singleImage) ? new StoredDataStatistics(total) : new Statistics();
final WeightedObservedPoints obs = new WeightedObservedPoints();
for (int i = (singleImage) ? 0 : start, j = 0; i < images.size(); i++) {
StringBuilder sb = (showTable) ? new StringBuilder() : null;
ImageSample sample = images.get(i);
for (PairSample pair : sample.samples) {
if (j % 16 == 0)
IJ.showProgress(j, total);
mean[j] = pair.getMean();
variance[j] = pair.variance;
// Gain is in ADU / e
double gain = variance[j] / (mean[j] - bias);
gainStats.add(gain);
obs.add(mean[j], variance[j]);
if (emMode) {
gain /= (2 * cameraGain);
}
if (showTable) {
sb.append(sample.title).append("\t");
sb.append(sample.exposure).append("\t");
sb.append(pair.slice1).append("\t");
sb.append(pair.slice2).append("\t");
sb.append(IJ.d2s(pair.mean1, 2)).append("\t");
sb.append(IJ.d2s(pair.mean2, 2)).append("\t");
sb.append(IJ.d2s(mean[j], 2)).append("\t");
sb.append(IJ.d2s(variance[j], 2)).append("\t");
sb.append(Utils.rounded(gain, 4)).append("\n");
}
j++;
}
if (showTable)
results.append(sb.toString());
}
IJ.showProgress(1);
if (singleImage) {
StoredDataStatistics stats = (StoredDataStatistics) gainStats;
Utils.log(TITLE);
if (emMode) {
double[] values = stats.getValues();
MathArrays.scaleInPlace(0.5, values);
stats = new StoredDataStatistics(values);
}
if (showCharts) {
// Plot the gain over time
String title = TITLE + " Gain vs Frame";
Plot2 plot = new Plot2(title, "Slice", "Gain", Utils.newArray(gainStats.getN(), 1, 1.0), stats.getValues());
PlotWindow pw = Utils.display(title, plot);
// Show a histogram
String label = String.format("Mean = %s, Median = %s", Utils.rounded(stats.getMean()), Utils.rounded(stats.getMedian()));
int id = Utils.showHistogram(TITLE, stats, "Gain", 0, 1, 100, true, label);
if (Utils.isNewWindow()) {
Point point = pw.getLocation();
point.x = pw.getLocation().x;
point.y += pw.getHeight();
WindowManager.getImage(id).getWindow().setLocation(point);
}
}
Utils.log("Single-image mode: %s camera", (emMode) ? "EM-CCD" : "Standard");
final double gain = stats.getMedian();
if (emMode) {
final double totalGain = gain;
final double emGain = totalGain / cameraGain;
Utils.log(" Gain = 1 / %s (ADU/e-)", Utils.rounded(cameraGain, 4));
Utils.log(" EM-Gain = %s", Utils.rounded(emGain, 4));
Utils.log(" Total Gain = %s (ADU/e-)", Utils.rounded(totalGain, 4));
} else {
cameraGain = gain;
Utils.log(" Gain = 1 / %s (ADU/e-)", Utils.rounded(cameraGain, 4));
}
} else {
IJ.showStatus("Computing fit");
// Sort
int[] indices = rank(mean);
mean = reorder(mean, indices);
variance = reorder(variance, indices);
// Compute optimal coefficients.
// a - b x
final double[] init = { 0, 1 / gainStats.getMean() };
final PolynomialCurveFitter fitter = PolynomialCurveFitter.create(2).withStartPoint(init);
final double[] best = fitter.fit(obs.toList());
// Construct the polynomial that best fits the data.
final PolynomialFunction fitted = new PolynomialFunction(best);
if (showCharts) {
// Plot mean verses variance. Gradient is gain in ADU/e.
String title = TITLE + " results";
Plot2 plot = new Plot2(title, "Mean", "Variance");
double[] xlimits = Maths.limits(mean);
double[] ylimits = Maths.limits(variance);
double xrange = (xlimits[1] - xlimits[0]) * 0.05;
if (xrange == 0)
xrange = 0.05;
double yrange = (ylimits[1] - ylimits[0]) * 0.05;
if (yrange == 0)
yrange = 0.05;
plot.setLimits(xlimits[0] - xrange, xlimits[1] + xrange, ylimits[0] - yrange, ylimits[1] + yrange);
plot.setColor(Color.blue);
plot.addPoints(mean, variance, Plot2.CROSS);
plot.setColor(Color.red);
plot.addPoints(new double[] { mean[0], mean[mean.length - 1] }, new double[] { fitted.value(mean[0]), fitted.value(mean[mean.length - 1]) }, Plot2.LINE);
Utils.display(title, plot);
}
final double avBiasNoise = Math.sqrt(noiseStats.getMean());
Utils.log(TITLE);
Utils.log(" Directory = %s", inputDirectory);
Utils.log(" Bias = %s +/- %s (ADU)", Utils.rounded(bias, 4), Utils.rounded(avBiasNoise, 4));
Utils.log(" Variance = %s + %s * mean", Utils.rounded(best[0], 4), Utils.rounded(best[1], 4));
if (emMode) {
final double emGain = best[1] / (2 * cameraGain);
// Noise is standard deviation of the bias image divided by the total gain (in ADU/e-)
final double totalGain = emGain * cameraGain;
Utils.log(" Read Noise = %s (e-) [%s (ADU)]", Utils.rounded(avBiasNoise / totalGain, 4), Utils.rounded(avBiasNoise, 4));
Utils.log(" Gain = 1 / %s (ADU/e-)", Utils.rounded(1 / cameraGain, 4));
Utils.log(" EM-Gain = %s", Utils.rounded(emGain, 4));
Utils.log(" Total Gain = %s (ADU/e-)", Utils.rounded(totalGain, 4));
} else {
// Noise is standard deviation of the bias image divided by the gain (in ADU/e-)
cameraGain = best[1];
final double readNoise = avBiasNoise / cameraGain;
Utils.log(" Read Noise = %s (e-) [%s (ADU)]", Utils.rounded(readNoise, 4), Utils.rounded(readNoise * cameraGain, 4));
Utils.log(" Gain = 1 / %s (ADU/e-)", Utils.rounded(1 / cameraGain, 4));
}
}
IJ.showStatus("");
}
Aggregations