public static double DistanceTo(this Recognizer.Centroid first, ImageInfo info)
{
double distance = 0.0;
for (int i = 0; i < ImageInfo.PropertiesCount; ++i)
{
distance += Math.Pow(first.PropertyValues[i] - info.GetPropertyByIndex(i), 2);
}
distance = Math.Sqrt(distance);
return distance;
}
public static List<ImageInfo> CalculateStats(int[,] pic, Dictionary<int, List<Point>> pointsByLabel)
{
Func<double, double> sqrt = Math.Sqrt;
Func<double, double> square = x => Math.Pow(x, 2);
int height = pic.GetLength(0);
int width = pic.GetLength(1);
var images = new List<ImageInfo>();
foreach (KeyValuePair<int, List<Point>> imagePoints in pointsByLabel)
{
var info = new ImageInfo();
info.Label = imagePoints.Key;
List<Point> points = imagePoints.Value;
foreach (Point p in points)
{
info.Area += 1;
if (p.X - 1 < 0 || p.X + 1 >= width || p.Y - 1 < 0 || p.Y + 1 >= height)
{ // border
info.Perimeter += 1;
}
else if (pic[p.Y, p.X + 1] == -1 || pic[p.Y, p.X - 1] == -1 ||
pic[p.Y + 1, p.X] == -1 || pic[p.Y - 1, p.X] == -1)
{
info.Perimeter += 1;
}
}
Debug.WriteLine("Area: {0}", info.Area);
Debug.WriteLine("Perimeter: {0}", info.Perimeter);
info.Compactness = square(info.Perimeter) / info.Area;
Debug.WriteLine("Compactness: {0}", info.Compactness);
info.XAverage = points.Sum(p => p.X) / info.Area;
info.YAverage = points.Sum(p => p.Y) / info.Area;
Debug.WriteLine("XAverage: {0}", info.XAverage);
Debug.WriteLine("YAverage: {0}", info.YAverage);
info.M20 = 0;
info.M11 = 0;
info.M02 = 0;
foreach (Point p in points)
{
info.M20 += (p.X - info.XAverage) * (p.X - info.XAverage);
info.M11 += (p.X - info.XAverage) * (p.Y - info.YAverage);
info.M02 += (p.Y - info.YAverage) * (p.Y - info.YAverage);
}
info.Elongation = (info.M20 + info.M02 + sqrt(square(info.M20 - info.M02) + 4 * info.M11 * info.M11))
/ (info.M20 + info.M02 - sqrt(square(info.M20 - info.M02) + 4 * info.M11 * info.M11));
Debug.WriteLine("Elongation: {0}", info.Elongation);
Debug.WriteLine("M11: " + info.M11);
Debug.WriteLine("M20: " + info.M20);
Debug.WriteLine("M02: " + info.M02);
info.Orientation = 0.5 * Math.Atan2(2 * info.M11, info.M20 - info.M02);
Debug.WriteLine("Orientation: " + info.Orientation);
images.Add(info);
}
foreach (ImageInfo image in images)
{
if (double.IsInfinity(image.Elongation))
{
Debugger.Break();
image.Elongation = images
.Where(im => !double.IsInfinity(im.Elongation))
.Max(im => im.Elongation);
}
else if (double.IsNaN(image.Elongation))
{
Debugger.Break();
}
}
return images;
}
private static int FindIndexOfNearestClusterMean(ImageInfo image, Centroid[] means)
{
Debug.WriteLine("Image's nearest mean:");
int nearestMeanIndex = -1;
double minDistance = double.MaxValue;
for (int i = 0; i < means.Length; ++i)
{
double distanceToMean = means[i].DistanceTo(image);
Debug.WriteLine("Distance to {0} = {1}", i, distanceToMean);
if (distanceToMean < minDistance)
{
nearestMeanIndex = i;
minDistance = distanceToMean;
}
}
if (nearestMeanIndex == -1) Debugger.Break();
Debug.WriteLine("Nearest mean index: {0}", nearestMeanIndex);
return nearestMeanIndex;
}
