private List <byte[]> get_palette(List <byte[]> samples1)
{
//def get_palette(samples, options, return_mask= False, kmeans_iter= 40):
//'''Extract the palette for the set of sampled RGB values. The first
//palette entry is always the background color; the rest are determined
//from foreground pixels by running K - means clustering.Returns the
//palette, as well as a mask corresponding to the foreground pixels.
var bg_color = get_bg_color(samples1, 6);
var fg_mask = get_fg_mask(bg_color, samples1);
//var kmeans_iter = 40; // Can't specify this at run time. 44 iterations are run versus 40 in python.
var kmeans = new Accord.MachineLearning.KMeans(num_colors - 1);
//Convert List of Byte[] to double[][] for Accord KMeans
var doubleSamples = new double[samples1.Count][];
for (int i = 0; i < samples1.Count; i++)
{
doubleSamples[i] = new double[] { samples1[i][0], samples1[i][1], samples1[i][2] }
}
;
// Filter samples to only true items
var countOfTrue = fg_mask.Where(x => x == true).Count();
var filteredSamples = new double[countOfTrue][];
var c = 0;
for (int i = 0; i < fg_mask.Count; i++)
{
if (fg_mask[i])
{
filteredSamples[c] = new double[] { doubleSamples[i][0], doubleSamples[i][1], doubleSamples[i][2] };
c++;
}
}
// Accord KMeans returns different values than scipy kmeans.
var clusters = kmeans.Learn(filteredSamples);
var palette1 = clusters.Centroids.ToList();
palette1.Insert(0, new double[] { bg_color.R, bg_color.G, bg_color.B });
var bytePal = new List <byte[]>();
for (int i = 0; i < palette1.Count; i++)
{
bytePal.Add(new byte[] { (byte)palette1[i][0], (byte)palette1[i][1], (byte)palette1[i][2] });
}
return(bytePal);
}
/// <summary>
/// Get the palette of an image
/// </summary>
/// <param name="image"></param>
public static List <Color> GetPalette(Image image, int paletteSize)
{
PaletteManager.PaletteSize = paletteSize;
// Get data for kmeans clustering
KMeansData data = GetKMeansData(image.Pixels.Select(p => p.Color).ToList());
// Format color data for KMeans (using Lab color space)
double[][] formattedData = FormatData(data.means);
double[][] centroids = FormatData(data.centroids);
// Perform KMeans w/additional black datapoint
Accord.MachineLearning.KMeans kMeans = new Accord.MachineLearning.KMeans(PaletteSize + 1);
kMeans.Centroids = centroids;
var clusters = kMeans.Learn(formattedData);
int[] labels = clusters.Decide(formattedData);
// Get palette values
double[][] doublePalette = kMeans.Centroids;
List <Color> Palette = new List <Color>();
foreach (double[] lab in doublePalette)
{
Palette.Add(new Color(L: lab[0], a: lab[1], b: lab[2]));
}
// Sort palette by luminance
Palette = Palette.OrderBy(p => p.L).ToList();
// Take all actual palette members
Palette = Palette.TakeLast(PaletteSize).ToList();
return(Palette);
}