private void btnGMMImage_Click(object sender, EventArgs e)
{
pic1.Load("Tennis Ball 620.jpg");
bmp = new Bitmap("Tennis Ball 620.jpg");
imageDataList = new List <ImagePoints>(bmp.Width * bmp.Height);
int k = int.Parse(txtNumClusters.Text);
int dim = 3;
imageDataList = InitializeImageData(bmp, k);
Matrix imgMatrix = new Matrix(bmp.Width * bmp.Height, dim);
for (int i = 0; i < (bmp.Width * bmp.Height); i++)
{
imgMatrix[i, 0] = imageDataList[i].red;
imgMatrix[i, 1] = imageDataList[i].green;
imgMatrix[i, 2] = imageDataList[i].blue;
}
GMM_NDim gmmnd = new GMM_NDim(k, dim, imgMatrix, bmp.Width * bmp.Height);
gmmnd.ComputeGMM_ND();
// determine class membership i.e., which point belongs to which cluster
imageDataList = new List <ImagePoints>();
for (int i = 0; i < imgMatrix.Rows; i++)
{
// Gamma matrix has the probabilities for a data point for its membership in each cluster
double[] probabs = new double[k];
int cnum = 0;
double maxprobab = gmmnd.Gamma[i, 0];
for (int m = 0; m < k; m++)
{
if (gmmnd.Gamma[i, m] > maxprobab)
{
cnum = m; // data i belongs to cluster m
maxprobab = gmmnd.Gamma[i, m];
}
}
ImagePoints imgpt = new ImagePoints {
clusterID = cnum, red = gmmnd.X[i, 0], green = gmmnd.X[i, 1], blue = gmmnd.X[i, 2]
};
imageDataList.Add(imgpt);
}
//MyImageProc.DrawClusters(pic1, imageDataList, 1, k);
//redraw image from rgb values
MessageBox.Show("img list len = " + imageDataList.Count.ToString() + " img width = " + bmp.Width.ToString() + " img height = " + bmp.Height.ToString());
//MessageBox.Show("new image done");
}
List <ImagePoints> InitializeImageData(Bitmap bmp, int numClusters)
{
System.Diagnostics.Debug.WriteLine("in intialise");
List <ImagePoints> PList = new List <ImagePoints>();
PList.Clear();
Random rand = new Random();
int dataLength = bmp.Width * bmp.Height; // number of data points
// create numCLuster distributions with different means and std devs
double[] meanX = new double[numClusters];
double[] meanY = new double[numClusters];
double[] meanZ = new double[numClusters];
double meanSampleX = 0;
double meanSampleY = 0;
double meanSampleZ = 0;
double[] stdDevX = new double[numClusters];
double[] stdDevY = new double[numClusters];
double[] stdDevZ = new double[numClusters];
double stdDevSampleX = 0;
double stdDevSampleY = 0;
double stdDevSampleZ = 0;
//initialise datapoints to RGB values from image
for (int i = 0; i < bmp.Width; i++)
{
for (int j = 0; j < bmp.Height; j++)
{
ImagePoints imgpt = new ImagePoints();
imgpt.red = (double)bmp.GetPixel(i, j).R;
imgpt.green = (double)bmp.GetPixel(i, j).G;
imgpt.blue = (double)bmp.GetPixel(i, j).B;
PList.Add(imgpt);
}
}
//selecting means i.e. random pts from the dataset.
//for (int i = 0; i < numClusters; i++)
//{
// meanX[i] = PList[rand.Next(0, dataLength)].red;
// meanY[i] = PList[rand.Next(0, dataLength)].green;
// meanZ[i] = PList[rand.Next(0, dataLength)].blue;
//}
////calculate sample mean
//for(int i = 0; i < dataLength; i++)
//{
// meanSampleX += PList[i].red;
// meanSampleY += PList[i].green;
// meanSampleZ += PList[i].blue;
//}
//meanSampleX /= dataLength;
//meanSampleY /= dataLength;
//meanSampleZ /= dataLength;
////calculate sample variance
//for(int i = 0; i < dataLength; i++)
//{
// stdDevSampleX += ((PList[i].red - meanSampleX) * (PList[i].red - meanSampleX));
// stdDevSampleY += ((PList[i].green - meanSampleY) * (PList[i].green - meanSampleY));
// stdDevSampleZ += ((PList[i].blue - meanSampleZ) * (PList[i].blue - meanSampleY));
//}
//stdDevSampleX /= dataLength;
//stdDevSampleY /= dataLength;
//stdDevSampleZ /= dataLength;
////set component variances to sample variance
//for (int i = 0; i < numClusters; i++)
//{
// stdDevX[i] = Math.Sqrt(stdDevSampleX);
// stdDevY[i] = Math.Sqrt(stdDevSampleY);
// stdDevZ[i] = Math.Sqrt(stdDevSampleZ);
//}
return(PList);
}
