private string GetColor(byte[] image)
{
// Lock Bits would be quicker
TypeConverter tc = TypeDescriptor.GetConverter(typeof(Bitmap));
Bitmap bitmap = (Bitmap)tc.ConvertFrom(image);
using (bitmap)
{
List <Color> colors = new List <Color>(bitmap.Width * bitmap.Height);
for (int x = 0; x < bitmap.Width; x++)
{
for (int y = 0; y < bitmap.Height; y++)
{
colors.Add(bitmap.GetPixel(x, y));
}
}
List <Color> filtered = colors
.Where(c => (KCluster.EuclideanDistance(c, Color.Black) >= 200) && (KCluster.EuclideanDistance(c, Color.White) >= 200))
.ToList();
KMeansClusteringCalculator clustering = new KMeansClusteringCalculator();
IList <Color> dominantColours = clustering.Calculate(10, filtered, 20.0d);
int index = closestColor(dominantColours.Skip(1).ToList(), dominantColours.First());
if (index == 0)
{
return(ToRGBAString(dominantColours[2]));
}
return(ToRGBAString(dominantColours[1]));
}
}
/// <summary>
/// Calculates the <paramref name="k"/> clusters for <paramref name="colours"/>. Iterations continues until clusters move by less than <paramref name="threshold"/>
/// </summary>
/// <param name="k">The number of clusters to calculate (eg. The number of results to return)</param>
/// <param name="colours">The list of colours to calculate <paramref name="k"/> for</param>
/// <param name="threshold">Threshold for iteration. A lower value should produce more correct results but requires more iterations and for some <paramref name="colours"/> may never produce a result</param>
/// <returns>The <paramref name="k"/> colours for the image in descending order from most common to least common</returns>
public IList <Color> Calculate(int k, IList <Color> colours, double threshold = 0.0d)
{
List <KCluster> clusters = new List <KCluster>();
// 1. Initialisation.
// Create K clusters with a random data point from our input.
// We make sure not to use the same index twice for two inputs
Random random = new Random();
List <int> usedIndexes = new List <int>();
while (clusters.Count < k)
{
int index = random.Next(0, colours.Count);
if (usedIndexes.Contains(index) == true)
{
continue;
}
usedIndexes.Add(index);
KCluster cluster = new KCluster(colours[index]);
clusters.Add(cluster);
}
bool updated = false;
do
{
updated = false;
// 2. For each colour in our input determine which cluster's centre point is the closest and add the colour to the cluster
foreach (Color colour in colours)
{
double shortestDistance = float.MaxValue;
KCluster closestCluster = null;
foreach (KCluster cluster in clusters)
{
double distance = cluster.DistanceFromCentre(colour);
if (distance < shortestDistance)
{
shortestDistance = distance;
closestCluster = cluster;
}
}
closestCluster.Add(colour);
}
// 3. Recalculate the clusters centre.
foreach (KCluster cluster in clusters)
{
if (cluster.RecalculateCentre(threshold) == true)
{
updated = true;
}
}
// 4. If we updated any centre point this iteration then iterate again
} while (updated == true);
return(clusters.OrderByDescending(c => c.PriorCount).Select(c => c.Centre).ToList());
}