public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TScan scan = inputArgs[0] as TScan;
int offset = (int) options[0].Value;
int width = (int) scan.Size.Width;
int height = (int) scan.Size.Height;
float[] data = scan.Data;
double quadrat_sum = 0.0;
unsafe {
fixed (float* ptrData = data) {
float* src = ptrData;
for (int i = 0; i < width * height; i++) {
quadrat_sum += *src;
src++;
}
}
}
double[] histogram = new double[offset];
for (int o = 0; o < offset; o++) {
for (int y = 0; y < height - o; y++) {
for (int x = 0; x < width - o; x++) {
histogram[o] += data[y * width + x] * data[(y + o) * width + (x + o)];
}
}
histogram[o] /= quadrat_sum;
}
return new IType[] { new THistogram(histogram) };
}
public unsafe override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TScan tScan = inputArgs[0] as TScan;
float[] data = tScan.Data;
int width = (int) tScan.Size.Width;
int height = (int) tScan.Size.Height;
float max = data.Max();
double M00 = 0, M10 = 0, M01 = 0;
double M11 = 0, M20 = 0, M02 = 0;
double M12 = 0, M21 = 0;
double M30 = 0, M03 = 0;
double InvM00 = 0;
double CenterX = 0, CenterY = 0;
for (int y = 0; y < height; y++) {
double yNorm = (double)y / height;
for (int x = 0; x < width; x++) {
double v = data[y * width + x] / max;
double xNorm = (double)x / width;
M00 += v;
M01 += yNorm * v;
M10 += xNorm * v;
M11 += xNorm * yNorm * v;
M02 += yNorm * yNorm * v;
M20 += xNorm * xNorm * v;
M21 += xNorm * xNorm * yNorm * v;
M12 += xNorm * yNorm * yNorm * v;
M30 += xNorm * xNorm * xNorm * v;
M03 += yNorm * yNorm * yNorm * v;
}
}
InvM00 = 1f / M00;
CenterX = M10 * InvM00;
CenterY = M01 * InvM00;
return new IType[] {
new TFeatureList<double>()
.AddFeature("M00", M00)
.AddFeature("M01", M01)
.AddFeature("M10", M10)
.AddFeature("M11", M11)
.AddFeature("M02", M02)
.AddFeature("M20", M20)
.AddFeature("M21", M21)
.AddFeature("M12", M12)
.AddFeature("M30", M30)
.AddFeature("M03", M03)
.AddFeature("CenterX", CenterX)
.AddFeature("CenterY", CenterY)
};
}
public unsafe override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
int bpp = (int) options[0].Value;
int p = (int) Math.Pow(2, bpp);
TScan tScan = inputArgs[0] as TScan;
float[] data = tScan.DataAs(bpp);
int height = (int) tScan.Size.Height;
int width = (int) tScan.Size.Width;
float[,] matrix = new float[p, width+1];
int maxRunLength = 0;
for (int y = 0; y < height; y++) {
int runLength = 1;
for (int x = 1; x < width; x++) {
int a = (int) data[width * y + (x - 1)];
int b = (int) data[width * y + x];
if (Math.Abs(a - b) <= float.Epsilon)
runLength++;
else {
matrix[a, runLength]++;
if (runLength > maxRunLength) {
maxRunLength = runLength;
}
runLength = 1;
}
if ((Math.Abs(a - b) <= float.Epsilon) && (x == width - 1)) {
matrix[a, runLength]++;
if (runLength > maxRunLength) {
maxRunLength = runLength;
}
}
if ((Math.Abs(a - b) > float.Epsilon) && (x == width - 1)) {
matrix[b, 1]++;
}
}
}
//if (crop) { // make this an option?
if (maxRunLength < width+1) {
float[,] matrixTmp = new float[p, maxRunLength+1];
for (int y = 0; y < maxRunLength+1; y++ ) {
for (int x = 0; x < p; x++) {
matrixTmp[x, y] = matrix[x, y];
}
}
matrix = matrixTmp;
}
//}
IType[] ret = { new TMatrix(matrix) };
return ret;
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TScan scan = inputArgs[0] as TScan;
int bpp = (int) options[0].Value;
int dx = (int) options[1].Value;
int dy = (int) options[2].Value;
int p = (int) Math.Pow(2, bpp);
TMatrix matrix;
if (bpp > 10) {
matrix = new TMatrix(new SparseMatrix<float>(p + 1, p + 1));
} else {
matrix = new TMatrix(new float[p + 1, p + 1]);
}
int width = (int) scan.Size.Width;
int height = (int) scan.Size.Height;
int startX = Math.Max(0, -dx);
int startY = Math.Max(0, -dy);
int endX = width - Math.Max(0, dx);
int endY = height - Math.Max(0, dy);
// float[] data = scan.Data;
// double max = data.Max();
//
// int offset = Math.Max(0, dx);
// if (max > 0.0) {
// unsafe {
// fixed (float* ptrData = data) {
// float* src = ptrData;
//
// int oldProgress = 0;
// for (int y = startY; y < endY; y++) {
// for (int x = startX; x < endX; x++, src++) {
// int posWithOffset = ((y + dy) * width) + (x + dx);
//
// int fromValue = (int) ((*src / max) * p);
// int toValue = (int) ((data[posWithOffset] / max) * p);
// matrix[fromValue, toValue] += increment;
//
// }
// src += offset;
//
// int progress = (int) (y * 100.0) / height;
// if (progress - oldProgress > 10) {
// oldProgress = progress;
// SetProgress(progress);
// }
// }
// }
// }
// }
IType[] ret = { matrix };
return ret;
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TScan scan = inputArgs[0] as TScan;
byte[] scanData = scan.DataAs8bpp();
int width = scan.Width;
int height = scan.Height;
return null;
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
IType input = inputArgs[0];
if (input.GetType() == typeof(TMatrix)) {
TMatrix matrix = input as TMatrix;
Console.WriteLine(matrix);
}
return null;
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
THistogram tHistogram = inputArgs[0] as THistogram;
double[] histogram = tHistogram.Data;
double mean = histogram.Mean();
double s = histogram.StandardDeviation(mean);
double skewness = 0.0;
double kurtosis = 0.0;
int lastValley = -1;
int lastHill = -1;
int hillCount = histogram[0] - histogram[1] > 0 ? 1 : 0;
for (int i = 0; i < histogram.Length-1; i++) {
skewness += Math.Pow(histogram[i] - mean, 3);
kurtosis += Math.Pow(histogram[i] - mean, 4);
double diff = histogram[i] - histogram[i + 1];
if (diff > 0 && lastValley == -1) { // new valley begins here
lastValley = i;
lastHill = -1; // finished hill
} else if (diff < 0 && lastHill == -1) { // new hill begins here
hillCount++;
lastHill = i;
if (lastValley != -1) { // there was a valley before
lastValley = -1; // finished valley
}
}
}
skewness += Math.Pow(histogram[histogram.Length-1] - mean, 3);
kurtosis += Math.Pow(histogram[histogram.Length-1] - mean, 4);
skewness /= (histogram.Length - 1) * s * s * s;
kurtosis /= (histogram.Length - 1) * s * s * s * s;
double regularity = (double) hillCount / (histogram.Length / 2);
return new IType[] {
new TFeatureList<double>()
.AddFeature("regularity", regularity)
.AddFeature("skewness", skewness)
.AddFeature("kurtosis", kurtosis)
};
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
int blockWidth = (int) options[0].Value;
int blockHeight = (int) options[1].Value;
Xwt.Size blockSize = new Xwt.Size(blockWidth, blockHeight);
TScan scan = inputArgs[0] as TScan;
int width = (int) scan.Size.Width;
int height = (int) scan.Size.Height;
float[] inputData = scan.Data;
for (int y = 0; y < height-blockHeight; y += blockHeight) {
for (int x = 0; x < width-blockWidth; x+= blockWidth) {
float[] data = new float[blockWidth*blockHeight];
bool empty = true;
for (int i = 0; i < blockWidth; i++) {
for (int j = 0; j < blockHeight; j++) {
float value = inputData[x + i + ((y + j) * width)];
data[i + (j * blockWidth)] = value;
if (value > 0.0) {
empty = false;
}
}
}
if (!empty) {
TScan block = new TScan(data, blockSize, scan.IsMultipleAccessModeOn);
Yield(new IType[] { block }, null);
}
}
}
return null;
}
/// <summary> /// Executes the algorithm. /// </summary> /// <param name="requestedData">Requested data.</param> /// <param name="options">User setted options</param> /// <param name="inputArgs">Input arguments.</param> /// <remarks> /// Return null, when no more data is available (important for sequential data output). /// Use Yield() function to return data when you want to output more then one result per input. /// </remarks> public abstract IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs);
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TMatrix tMatrix = inputArgs[0] as TMatrix;
if (tMatrix == null) {
return null;
}
HaralickIntern h = new HaralickIntern(tMatrix);
// features
double asm = 0.0, contrast = 0.0, correlation, variance = 0.0, homogeneity = 0.0;
double sumAverage = 0.0, sumVariance = 0.0, sumEntropy = 0.0, entropy = 0.0;
double diffVariance = 0.0, diffEntropy = 0.0;
double informationMeasure1 = 0.0, informationMeasure2 = 0.0;
// temp variables
double correlationStep = 0.0;
double informationMeasure1Step = 0.0, informationMeasure2Step = 0.0;
double[] xydiff = new double[Math.Max(h.InputMatrix.Width, h.InputMatrix.Height)];
double mean = h.InputMatrix.Mean;
// compute
//Parallel.For(0, h.InputMatrix.GetLength(0), i => {
for (int i = 0; i < h.InputMatrix.Width; i++) {
for (int j = 0; j < h.InputMatrix.Height; j++) {
double value = h.InputMatrix[i, j];
xydiff[Math.Abs(i - j)] += h.InputMatrix[i, j];
correlationStep += (i * j) * h.InputMatrix[i, j];
informationMeasure1Step -= h.InputMatrix[i, j] * Math.Log(h.Px[i] * h.Py[j] + double.Epsilon, 2);
informationMeasure2Step -= h.Px[i] * h.Py[j] * Math.Log(h.Px[i] * h.Py[j] + double.Epsilon, 2);
asm += value * value;
variance += (i - mean) * h.InputMatrix[i, j];
homogeneity += h.InputMatrix[i, j] / (double) (1 + (i - j) * (i - j));
}
} //);
correlation = (correlationStep - h.MeanX * h.MeanY) / (h.StandardDeviationX * h.StandardDeviationY);
for (int n = 0; n < xydiff.Length; n++) {
contrast += (n * n) * xydiff[n];
}
SetProgress(75);
sumEntropy = h.Sums.Entropy(double.Epsilon);
for (int i = 0; i < h.Sums.Length; i++) {
sumAverage += i * h.Sums[i];
sumVariance += (i - sumEntropy) * (i - sumEntropy) * h.Sums[i];
}
entropy = h.InputMatrix.Entropy(float.Epsilon);
diffVariance = xydiff.Variance();
diffEntropy = xydiff.Entropy(double.Epsilon);
double entropyX = h.Px.Entropy(double.Epsilon);
double entropyY = h.Py.Entropy(double.Epsilon);
informationMeasure1 = (entropy - informationMeasure1Step) / Math.Max(entropyX, entropyY);
informationMeasure2 = Math.Pow(1.0 - Math.Exp(-2 * (informationMeasure2Step - entropy)), 0.5);
return new IType[] {
new TFeatureList<double>()
.AddFeature("asm", asm)
.AddFeature("contrast", contrast)
.AddFeature("correlation", correlation)
.AddFeature("variance", variance)
.AddFeature("homogeneity", homogeneity)
.AddFeature("sumAverage", sumAverage)
.AddFeature("sumVariance", sumVariance)
.AddFeature("sumEntropy", sumEntropy)
.AddFeature("entropy", entropy)
.AddFeature("diffVariance", diffVariance)
.AddFeature("diffEntropy", diffEntropy)
.AddFeature("informationMeasure1", informationMeasure1)
.AddFeature("informationMeasure2", informationMeasure2)
};
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TScan tScan = inputArgs[0] as TScan;
bool normalize = (bool) options[0].Value;
float[] data = tScan.Data;
int width = tScan.Width;
int height = tScan.Height;
if (normalize) {
float mean = data.Mean();
for (int y = 0; y < height; y++) {
for (int x = 0; x < width; x++) {
data[(y * width) + x] /= mean;
}
}
}
float[,] l5e5d = new float[(width - 3), (height - 3)];
float[,] l5s5d = new float[(width - 3), (height - 3)];
float[,] l5r5d = new float[(width - 3), (height - 3)];
float[,] e5e5d = new float[(width - 3), (height - 3)];
float[,] e5s5d = new float[(width - 3), (height - 3)];
float[,] e5r5d = new float[(width - 3), (height - 3)];
float[,] e5l5d = new float[(width - 3), (height - 3)];
float[,] s5l5d = new float[(width - 3), (height - 3)];
float[,] s5e5d = new float[(width - 3), (height - 3)];
float[,] s5r5d = new float[(width - 3), (height - 3)];
float[,] s5s5d = new float[(width - 3), (height - 3)];
float[,] r5r5d = new float[(width - 3), (height - 3)];
float[,] r5e5d = new float[(width - 3), (height - 3)];
float[,] r5s5d = new float[(width - 3), (height - 3)];
float[,] r5l5d = new float[(width - 3), (height - 3)];
for (int y = 2; y < height - 2; y++) {
for (int x = 2; x < width - 2; x++) {
for (int i = -2; i <= 2; i++) {
for (int j = -2; j <= 2; j++) {
l5e5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * l5e5[i + 2, j + 2];
l5s5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * l5s5[i + 2, j + 2];
l5r5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * l5r5[i + 2, j + 2];
e5e5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * e5e5[i + 2, j + 2];
e5s5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * e5s5[i + 2, j + 2];
e5r5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * e5r5[i + 2, j + 2];
e5l5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * e5l5[i + 2, j + 2];
s5l5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * s5l5[i + 2, j + 2];
s5e5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * s5e5[i + 2, j + 2];
s5r5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * s5r5[i + 2, j + 2];
s5s5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * s5s5[i + 2, j + 2];
r5r5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * r5r5[i + 2, j + 2];
r5e5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * r5e5[i + 2, j + 2];
r5s5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * r5s5[i + 2, j + 2];
r5l5d[x - 2, y - 2] += data[((y + j) * width) + x + i] * r5l5[i + 2, j + 2];
}
}
}
}
return new IType[] {
new TFeatureList<double>()
.AddFeature("L5E5/E5L5", l5e5d.AbsMean() / e5l5d.AbsMean())
.AddFeature("L5S5/S5L5", l5s5d.AbsMean() / s5l5d.AbsMean())
.AddFeature("L5R5/R5L5", l5r5d.AbsMean() / r5l5d.AbsMean())
.AddFeature("E5S5/S5E5", e5s5d.AbsMean() / s5e5d.AbsMean())
.AddFeature("E5R5/R5E5", e5r5d.AbsMean() / r5e5d.AbsMean())
.AddFeature("S5R5/R5S5", s5r5d.AbsMean() / r5s5d.AbsMean())
.AddFeature("E5E5", e5e5d.AbsMean())
.AddFeature("S5S5", s5s5d.AbsMean())
.AddFeature("R5R5", r5r5d.AbsMean())
};
}
public unsafe override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TScan scan = inputArgs[0] as TScan;
int blockSize = (int) Math.Pow(2, (int) options[0].Value);
bool normalize = (bool) options[1].Value;
bool rotationInvariant = (bool) options[2].Value;
bool uniformLBP = (bool) options[3].Value;
if (uniformLBP) {
rotationInvariant = true;
}
byte[] scanData = scan.DataAs8bpp();
int width = scan.Width;
int height = scan.Height;
int[] nPatterns = new int[height * width];
fixed (byte* ptrData = scanData) {
fixed (int* ptrNPatterns = nPatterns) {
// skip the first line
byte* data = ptrData + width;
int* neighbor = ptrNPatterns + width;
for (int y = 1; y < height - 1; y++) {
// and skip first column
data++;
neighbor++;
for (int x = 1; x < width - 1; x++, data++, neighbor++) {
byte n11 = *(data + width + 1);
byte n12 = *(data + 1);
byte n13 = *(data - width + 1);
byte n21 = *(data + width);
byte n22 = *data;
byte n23 = *(data - width);
byte n31 = *(data + width - 1);
byte n32 = *(data - 1);
byte n33 = *(data - width - 1);
int sum = 0;
if (n22 < n11)
sum += 1 << 0;
if (n22 < n12)
sum += 1 << 1;
if (n22 < n13)
sum += 1 << 2;
if (n22 < n21)
sum += 1 << 3;
if (n22 < n23)
sum += 1 << 4;
if (n22 < n31)
sum += 1 << 5;
if (n22 < n32)
sum += 1 << 6;
if (n22 < n33)
sum += 1 << 7;
if (rotationInvariant && sum != 0) {
while ((sum & 0x80) == 0) {
sum <<= 1;
}
}
*neighbor = sum;
}
neighbor++;
data++;
}
}
}
int numberOfBlocksH = width / blockSize;
int numberOfBlocksV = height / blockSize;
// compute histogram
double[] histogram;
if (uniformLBP) {
histogram = new double[6 * numberOfBlocksH * numberOfBlocksV];
int blockNumber = 0;
for (int y = 0; y < height; ) {
for (int x = 0; x < width; ) {
switch (nPatterns[x + y * width]) {
case 0: // flat
histogram[6 * blockNumber]++;
break;
case 255: // flat
histogram[1 + 6 * blockNumber]++;
break;
case 128: // line end
histogram[2 + 6 * blockNumber]++;
break;
case 224: // edge
histogram[3 + 6 * blockNumber]++;
break;
case 252: // corner
histogram[4 + 6 * blockNumber]++;
break;
default:
histogram[5 + 6 * blockNumber]++;
break;
}
x++;
if (x % blockSize == 0) {
blockNumber++;
}
}
y++;
if (y % blockSize != 0) {
blockNumber -= numberOfBlocksH;
}
}
} else {
histogram = new double[256];
for (int y = 0; y < height; y++) {
for (int x = 0; x < width; x++) {
histogram[nPatterns[x + y * width]]++;
}
}
histogram[0] = 0d;
}
double histogramMax = histogram.Max();
TFeatureList<double> featureList = new TFeatureList<double>();
for (int i = 0; i < histogram.Length; i++) {
if (normalize) {
histogram[i] /= histogramMax;
}
featureList.AddFeature("LBP#" + i, histogram[i]);
}
return new IType[] {
featureList,
new THistogram(histogram)
};
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TMatrix matrix = inputArgs[0] as TMatrix;
if (matrix == null) {
return null;
}
double numberOfRuns = matrix.Sum; // R_G
double sre = 0.0, lre = 0.0, lgre = 0.0, hlre = 0.0;
double gln = 0.0, rln = 0.0;
int n = 0; // number of pixels in original image
for (int i = 0; i < matrix.Width; i++) {
double glnInner = 0.0;
for (int j = 0; j < matrix.Height; j++) {
if (matrix[i, j] >= double.Epsilon) {
glnInner += matrix[i, j];
if (j > 0) {
n += (int) (matrix[i, j] * j);
sre += matrix[i, j] / (j * j);
lre += matrix[i, j] * (j * j);
}
if (i > 0) {
lgre += matrix[i, j] / (i * i);
hlre += matrix[i, j] * (i * i);
}
}
}
gln += glnInner * glnInner;
}
SetProgress(50);
for (int j = 0; j < matrix.Height; j++) {
double rlnInner = 0.0;
for (int i = 0; i < matrix.Width; i++) {
rlnInner += matrix[i, j];
}
rln += rlnInner * rlnInner;
}
sre /= numberOfRuns;
lre /= numberOfRuns;
gln /= numberOfRuns;
rln /= numberOfRuns;
double rp = numberOfRuns / n;
return new IType[] {
new TFeatureList<double>()
.AddFeature("shortRunEmphasis", sre)
.AddFeature("longRunEmphasis", lre)
.AddFeature("grayLevelNonUniformity", gln)
.AddFeature("runLengthNonUniformity", rln)
.AddFeature("runPercentage", rp)
.AddFeature("lowGrayLevelRunEmphasis", lgre)
.AddFeature("highGrayLevelRunEmphasis", hlre)
};
}
public unsafe override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
TScan tScan = inputArgs[0] as TScan;
float[] data = tScan.Data;
int nrOfBins = (int) options[0].Value;
int width = (int) tScan.Size.Width;
int height = (int) tScan.Size.Height;
double coarsness = 0.0, contrast = 0.0, directionality = 0.0;
float[,] sumAreaTable = SumAreaTable(data, width, height);
float mean = data.Mean();
float sigma = data.StandardDeviation(mean);
double moment4 = 0.0;
double[] histogram = new double[nrOfBins];
double binWindow = (histogram.Length - 1) / Math.PI;
int bin = -1;
int oldProgress = 0;
for (int y = 0; y < height; y++) {
for (int x = 0; x < width; x++) {
moment4 += Math.Pow(data[y * width + x] - mean, 4);
if (x > 0 && y > 0) {
coarsness += Math.Pow(2, Sopt(sumAreaTable, x, y));
if (x < width - 1 && y < height - 1) {
double v = DeltaV(data, width, x, y);
double h = DeltaH(data, width, x, y);
if (h > 0.0 && v > 0.0) {
bin = (int) ((Math.PI / 2 + Math.Atan(v / h)) * binWindow);
histogram[bin]++;
}
}
}
}
int progress = (int) (y * 100.0) / height;
if (progress - oldProgress > 5) {
oldProgress = progress;
SetProgress(progress);
}
}
coarsness /= height * width;
if (sigma > 0) { // sigma <= 0 only if image is complete black
double alpha4 = moment4 / (Math.Pow(sigma, 4));
contrast = sigma / Math.Pow(alpha4, 0.25);
}
double histSum = histogram.Sum();
for (int i = 0; i < histogram.Length; i++) {
histogram[i] /= histSum;
}
int lastValley = -1;
int lastHill = -1;
int hillCount = 0; // n_p
for (int i = 0; i < histogram.Length-1; i++) {
double diff = histogram[i] - histogram[i + 1];
if (diff > 0 && lastValley == -1) { // new valley begins here
lastValley = i;
lastHill = -1; // finished hill
} else if (diff < 0 && lastHill == -1) { // new hill begins here
hillCount++;
lastHill = i;
if (lastValley != -1) { // there was a valley before
for (int j = lastValley; j < i; j++) {
directionality += Math.Pow(j - i, 2) * (histogram[j]);
}
lastValley = -1; // finished valley
}
} else if (diff < 0) { // it still goes down
directionality += Math.Pow(i - lastHill, 2) * (histogram[i]);
}
}
directionality = 1 - (binWindow * hillCount * directionality);
return new IType[] {
new TFeatureList<double>()
.AddFeature("coarsness", coarsness)
.AddFeature("contrast", contrast)
.AddFeature("directionality", directionality),
new THistogram(histogram)
};
}
public override IType[] Run(Dictionary<RequestType, object> requestedData, BaseOption[] options, IType[] inputArgs)
{
ScanCollection scans = requestedData[RequestType.ScanCollection] as ScanCollection;
int size = scans.Count;
bool maskedOnly = (bool) options[0].Value;
int i = 0;
foreach (BaseScan scan in scans) {
if (IsCanceled) {
break;
}
IType[] data = new IType[1];
if ((maskedOnly && scan.HasMask) || !maskedOnly) {
if (scan.AvailableScanTypes().Contains(scanTypeComboBox.Value)) {
data[0] =
new TScan(scan, scanTypeComboBox.Value.ToString(), maskedOnly: maskedOnly).Preload();
Yield(data, scan);
}
}
i++;
SetProgress((i * 100) / size);
}
return null;
}
