// Incrementally blur the input image first so it reaches the next octave.
public void BuildGaussianMaps(ImageMap first, double firstScale,
int scales, double sigma)
{
// We need one more gaussian blurred image than the number of DoG
// maps. But for the minima/maxima pixel search, we need two more. See
// BuildDiffMaps.
imgScaled = new ImageMap[scales + 1 + 1 + 1];
this.basePixScale = firstScale;
// Convolve first image with the octaveSigma. Previously we got this
// wrong, but thanks to Alexandre Jenny, this got fixed. Thanks!
GaussianConvolution gauss = new GaussianConvolution(sigma);
ImageMap prev = imgScaled[0] = gauss.Convolve(first);
// Ln1(x, y, k^{p+1}) = G(x, y, k) * Ln0(x, y, k^p).
for (int scI = 1; scI < imgScaled.Length; ++scI)
{
gauss = new GaussianConvolution(sigma);
// TODO: real fix, "first" is the correct behaviour, however, a
// large sigma leads to a large convolution kernel -> slow
// better: incremental convolution with smaller sigma.
prev = imgScaled[scI] = gauss.Convolve(first);
sigma *= SToK(scales);
}
}
public void BuildGaussianMaps(ImageMap first, double firstScale,
int scales, double sigma)
{
imgScaled = new ImageMap[scales + 1 + 1 + 1];
this.basePixScale = firstScale;
ImageMap prev = first;
imgScaled[0] = first;
double w = sigma;
double kTerm = Math.Sqrt(Math.Pow(SToK(scales), 2.0) - 1.0);
for (int scI = 1; scI < imgScaled.Length; ++scI)
{
GaussianConvolution gauss = new GaussianConvolution(w * kTerm);
prev = imgScaled[scI] = gauss.Convolve(prev);
w *= SToK(scales);
}
#if false
for (int scI = 0; scI < imgScaled.Length; ++scI)
{
imgScaled[scI].Save("scale-" + scI + ".dat", null);
if (scI < (imgScaled.Length - 1))
{
(imgScaled[scI + 1] - imgScaled[scI]).Save
("scale-diff-" + scI + ".dat", null);
}
}
Console.WriteLine("DEBUG EXIT");
#endif
}
//OK, 04/13/2009
public void ConstructScaleSpace()
{
Matrix tempMatrix = IRI.Ket.DigitalImageProcessing.GeometricEnhancement.DoubleTheSize(this.originalImage);
tempMatrix = new GaussianConvolution(1.5).Convolve(tempMatrix);
originalStandardDeviation *= 0.5;
for (int n = 0; n < depthOfScale; n++)
{
//We need s + 3 images in each octave. [David Lowe 2004]
for (int s = 0; s < depthOfBlure + 3; s++)
{
if (s == 0)
{
//if (n != 0)
//{
scaleSpace.Add(tempMatrix);
//}
continue;
}
double standarDeviation = CalculateSigma(n, s);
scaleSpace.Add(new GaussianConvolution(standarDeviation).Convolve(tempMatrix));
}
// index of image in the current octave (has the twice
// scale of first image in the current octave) to be
// half,
int tempIndex = n * (this.depthOfBlure + 3) + depthOfBlure;
tempMatrix = IRI.Ket.DigitalImageProcessing.GeometricEnhancement.HalveTheSize(scaleSpace[tempIndex]);
}
}
public int DetectFeaturesDownscaled(ImageMap img, int bothDimHi,
double startScale)
{
globalKeypoints = globalNaturalKeypoints = null;
if (printWarning)
{
Console.Error.WriteLine("");
Console.Error.WriteLine("===============================================================================");
Console.Error.WriteLine("The use of this software is restricted by certain conditions.");
Console.Error.WriteLine("See the \"LICENSE\" file distributed with the program for details.");
Console.Error.WriteLine("");
Console.Error.WriteLine("The University of British Columbia has applied for a patent on the SIFT");
Console.Error.WriteLine("algorithm in the United States. Commercial applications of this software may");
Console.Error.WriteLine("require a license from the University of British Columbia.");
Console.Error.WriteLine("===============================================================================");
Console.Error.WriteLine("");
}
if (bothDimHi < 0)
{
img = img.ScaleDouble();
startScale *= 0.5;
}
else if (bothDimHi > 0)
{
while (img.XDim > bothDimHi || img.YDim > bothDimHi)
{
img = img.ScaleHalf(); //pass
startScale *= 2.0;
}
}
if (preprocSigma > 0.0)
{
GaussianConvolution gaussianPre =
new GaussianConvolution(preprocSigma);
img = gaussianPre.Convolve(img);
}
pyr = new OctavePyramid();
pyr.Verbose = verbose;
pyr.BuildOctaves(img, startScale, scaleSpaceLevels,
octaveSigma, minimumRequiredPixelsize);
globalKeypoints = new ArrayList();
for (int on = 0; on < pyr.Count; ++on)
{
DScaleSpace dsp = pyr[on];
ArrayList peaks = dsp.FindPeaks(dogThresh);
if (verbose)
{
Console.WriteLine("Octave {0} has {1} raw peaks",
on, peaks.Count);
}
int oldCount = peaks.Count;
ArrayList peaksFilt = dsp.FilterAndLocalizePeaks(peaks,
maximumEdgeRatio, dValueLowThresh, scaleAdjustThresh,
relocationMaximum);
if (verbose)
{
Console.WriteLine(" filtered: {0} remaining from {1}, thats % {2:N2}",
peaksFilt.Count, oldCount, (100.0 * peaksFilt.Count) / oldCount);
Console.WriteLine("generating keypoints from peaks");
}
dsp.GenerateMagnitudeAndDirectionMaps();
ArrayList keypoints = dsp.GenerateKeypoints(peaksFilt,
scaleSpaceLevels, octaveSigma);
dsp.ClearMagnitudeAndDirectionMaps();
globalKeypoints.AddRange(keypoints);
}
return(globalKeypoints.Count);
}