public static double[] GradientTowardsMinimalErrorForGaussian(double peak, double sigma, ushort[] samples, double center, GaussianFitOptions2 options)
{
var vec = new double[2];
var used = 0;
for (var x = 0; x < samples.Length; x++)
{
var distance = (x - center);
if (distance <= 5)
{
continue;
}
var distanceSquared = distance * distance;
var sample = samples[x];
var sChange = Photometry2.SigmaSlopeFromDerivative((double)sample, (double)x, center, sigma, peak);
var pChange = Photometry2.PeakSlopeFromDerivative((double)sample, (double)x, center, sigma, peak);
if (double.IsPositiveInfinity(sChange) || double.IsNegativeInfinity(sChange))
{
throw new ArgumentException();
}
if (double.IsPositiveInfinity(pChange) || double.IsNegativeInfinity(pChange))
{
throw new ArgumentException();
}
vec[0] += (1 * sChange);
vec[1] += (1 * pChange);
used++;
}
vec[0] /= used;
vec[1] /= used;
return(vec);
}
public static double[] GradientTowardsMinimalErrorForGaussian(double peak, double sigma, ushort[] samples, double center, GaussianFitOptions2 options)
{
var vec = new double[2];
var used = 0;
for (var x = 0; x < samples.Length; x++)
{
var distance = (x - center);
if (distance > 10)
{
continue;
}
var distanceSquared = distance * distance;
if (distanceSquared == 0)
{
continue;
}
var sample = samples[x];
if (sample == peak)
{
continue;
}
var sigmaSample = GaussianSigmaFromSample(peak, distanceSquared, sample);
var peakSample = GaussianPeakFromSample(distanceSquared, sigma, sample);
var sChange = GaussianSigmaErrorSlopeFunction(distanceSquared, peak, sigmaSample, sigma);
var pChange = GaussianPeakErrorSlopeFunction(peakSample, peak, sigma);
if (double.IsPositiveInfinity(sChange) || double.IsNegativeInfinity(sChange))
{
throw new ArgumentException();
}
vec[0] += (-1 * sChange);
vec[1] += (-1 * pChange);
used++;
}
vec[0] /= used;
vec[1] /= used;
return(vec);
}
public static GaussianFit FindGaussianPSF(ushort[] samples, double peak, double center, GaussianFitOptions2 options)
{
var width = options.StartSigma;
var widthStep = 1.0;
var peakStep = 1.0;
var iterations = 0;
for (iterations = 0; iterations < options.MaxIterations && Math.Abs(widthStep) > options.MinimumChangeThreshold; iterations++)
{
var direction = GradientTowardsMinimalErrorForGaussian(peak, width, samples, center, options);
widthStep = options.IterationStepSize * direction[0];
width = width + widthStep;
//peakStep = options.IterationStepSize * direction[1];
//peak = peak + peakStep;
if (width < .1)
{
Debug.WriteLine($"Error: sigma got too small");
return(new GaussianFit {
Result = GaussianFitResult.Error
});
}
else if (width > 20)
{
Debug.WriteLine($"Error: sigma got too large");
return(new GaussianFit {
Result = GaussianFitResult.Error
});
}
if (peak < 0)
{
Debug.WriteLine($"Error: peak got too small");
return(new GaussianFit {
Result = GaussianFitResult.Error
});
}
PrintResiduals(samples, peak, center, width);
}
return(new GaussianFit
{
Width = width,
Peak = peak,
Iterations = iterations,
Result = iterations == options.MaxIterations ? GaussianFitResult.IterationsMaxed : GaussianFitResult.StepMinimumReached
});
}
public static GaussianFit FindGaussianPSF(ushort[] samples, double peak, double center, GaussianFitOptions2 options)
{
var width = 1.0;
var widthStep = 1.0;
var iterations = 0;
for (iterations = 0; iterations < options.MaxIterations && Math.Abs(widthStep) > options.MinimumChangeThreshold; iterations++)
{
var direction = GradientTowardsMinimalErrorForGaussian(peak, width, samples, center, options);
widthStep = options.IterationStepSize * direction[0];
width = width + widthStep;
if (width < .1)
{
return(new GaussianFit {
Result = GaussianFitResult.Error
});
}
else if (width > 20)
{
return(new GaussianFit {
Result = GaussianFitResult.Error
});
}
if (peak < 0)
{
return(new GaussianFit {
Result = GaussianFitResult.Error
});
}
}
return(new GaussianFit
{
Width = width,
Peak = peak,
Iterations = iterations,
Result = iterations == options.MaxIterations ? GaussianFitResult.IterationsMaxed : GaussianFitResult.StepMinimumReached
});
}
