public static UBitmap FromImage(Image image)
{
var bmp = new Bitmap(image);
var ubmp = new UBitmap(image.Width, image.Height);
for (var y = 0; y < image.Height - 1; y++)
{
for (var x = 0; x < image.Width - 1; x++)
{
ubmp.SetPixel(x, y, bmp.GetPixel(x, y));
}
}
return(ubmp);
}
/// <summary>
/// Detect objects in the image
/// </summary>
/// <param name="image"></param>
/// <returns></returns>
public static Detection[] DetectObjects(this Image image, float accuracy = 20.0f)
{
//Lower the size of the image so that it takes less resources and less time
//The factor by which to downscale the size, determined by the accuracy given
var resizeScale = image.Width / ((image.Width / 100) * accuracy);
resizeScale = (image.Width / resizeScale) % 2 == 0 ? resizeScale : (int)(resizeScale + 1);
//The width we want on the resized image
var targetWidth = image.Width / resizeScale;
//resize the image
image = image.Resize((int)targetWidth);
//Detect edges first so that it becomes easier to distinguish one object from another
var ubmp = UBitmap.FromImage(image.DetectEdges());
//We'll use this list to keep our 'Points' of interest (no pun intended ;-) )
var detections = new List <Point>();
//Iterate through each pixel in the image row by row
//y is row
for (var y = 0; y < ubmp.Height - 1; y++)
{
//x is column
for (var x = 0; x < ubmp.Width - 1; x++)
{
//Get the pixel represented at the (x,y) coordinates
var pixelColor = ubmp.GetPixel(x, y);
//get the argb color code of the pixel
var pixel = pixelColor.ToArgb();
//check if pixel is white
if (pixel != -1)
{
//if not white, add the coordinates to the list
detections.Add(new Point(x, y));
}
}
}
//If there are no points of interest, we can assume that we have no objects in the image, so no need to continue ¯\_(ツ)_/¯
if (!detections?.Any() ?? true)
{
return(null);
}
//if multiple points are 'clumped' together we can assume that its from the same object
var clumps = new List <List <Point> >();
//add the first point as the first point of the first clump
clumps.Add(new List <Point> {
detections.First()
});
//Always keeping a record of what was the last point we processed
var lastPoint = clumps.Last().Last();
//iterate over each point except the first one since we added it already
foreach (var point in detections.Skip(1))
{
//The x coordinates of the last point and this point
var xs = new int[] { lastPoint.X, point.X };
//The y coordinates of the last point and this point
var ys = new int[] { lastPoint.Y, point.Y };
//the difference in the lastpoint's x coordinates
var xDiff = xs.Max() - xs.Min();
//the difference in the lastpoint's y coordinates
var yDiff = ys.Max() - ys.Min();
//Check to see if the two points are 'clumped together' (there is no blank-white-space between them)
//This tells us that the points are of the same object
if (xDiff < 2 || yDiff < 2)
{
//see if we have a previous record of the lastPoint (if lastpoint was interesting and of the same object)
var clump = clumps.FirstOrDefault(c => c.Any(p => p.Equals(lastPoint)));
//if no record exists, add a new record because why not?
if (clump == null)
{
clump = new List <Point>();
//add the new record to the list
clumps.Add(clump);
}
//add the point to the record
clump.Add(point);
//update the lastpoint
lastPoint = point;
}
//this block would be processed if there is atleast one blank-white-space between this point and the one before
else
{
//create a new record, since this is the first time we are dealing with this object
var clump = new List <Point>();
//add the record to the list
clumps.Add(clump);
//add the point to the record
clump.Add(point);
//update the lastpoint
lastPoint = point;
}
}
//Make Detection objects out of our clumps (convert them into something meaninful and interpretable)
//In here we multiply every X and Y values with resizeScale because the current points are relative to our resized image,
//But we need to know where these points were on the original full-size image
//We do Math.Ceiling because we cannot expect the value which comes after multiplying by resizeScale is gonna be a whole number,
//in which case we may lose a few pixels
var objects = clumps.Select((clump, i) =>
{
//Create a new Detection Object
var d = new Detection
{
//Use index as Id
Id = i,
//find lowest X coordinate in the clump
LX = (int)Math.Ceiling(clump.Min(x => x.X) * resizeScale),
//find lowest y coordinate in the clump
LY = (int)Math.Ceiling(clump.Min(y => y.Y) * resizeScale),
//find highest X coordinate in the clump
HX = (int)Math.Ceiling(clump.Max(x => x.X) * resizeScale),
//find highest X coordinate in the clump
HY = (int)Math.Ceiling(clump.Max(y => y.Y) * resizeScale)
};
return(d);
}).ToArray();
//return the objects;
return(objects);
//var target = new Bitmap(image);
//using (var pen = new Pen(Color.Aqua))
//using(var g = Graphics.FromImage(target))
//foreach (var obj in objects)
//{
// g.DrawRectangle(pen, obj.Rectangle);
//}
//return target;
}
/// <summary>
/// Detect the edges in the image
/// </summary>
/// <param name="image"></param>
/// <returns></returns>
public static Image DetectEdges(this Image image)
{
//Downscale or Upscale
//image = image.Resize(500);
//image = image.MakeGrayscale3();
var bmp = UBitmap.FromImage(image);
var target = new UBitmap(bmp.Width, bmp.Height);
//var pixels = new List<Point>();
//Iterate through each pixel in the image row by row
//y is row
for (var y = 0; y < bmp.Height - 1; y++)
{
//x is column
for (var x = 0; x < bmp.Width - 1; x++)
{
//Get the pixel represented at the (x,y) coordinates
var pixelColor = bmp.GetPixel(x, y);
//get the argb color code of the pixel
var pixel = pixelColor.ToArgb();
//make sure the pixel is not the first one on either of X and Y planes, because we need to process neighbouring pixels
if (x != 0 && y != 0)
{
//Pixel directly above this one, the names should be self explanatory, so not going into details of those.
var top = bmp.GetPixel(x, y - 1).ToArgb();
var left = bmp.GetPixel(x - 1, y).ToArgb();
var bottom = bmp.GetPixel(x, y + 1).ToArgb();
var right = bmp.GetPixel(x + 1, y).ToArgb();
var topLeft = bmp.GetPixel(x - 1, y - 1).ToArgb();
var topRight = bmp.GetPixel(x + 1, y - 1).ToArgb();
var bottomLeft = bmp.GetPixel(x - 1, y + 1).ToArgb();
var bottomRight = bmp.GetPixel(x + 1, y + 1).ToArgb();
//Take all of the neighbouring pixels into an array for easier processing
var colors = new int[] { top, left, bottom, right, topLeft, topRight, bottomLeft, bottomRight };
//check whether this pixel is worthy of being added to the 'edge' detection
var shouldAdd = colors.Select(color =>
{
//the pixels to compare
//pixel is this pixel, color is the neighbouring one
var compare = new int[] { pixel, color };
//Find the difference between the two
var difference = compare.Max() - compare.Min();
//Find the percentage difference between the two
//Also we need a positive number, whether they turn negative or positive,
//so we convert any negatives into positives and leave positives untouched
//If that didn't make any sense, you should read about 'Math.Abs' (Absolute) function
//There should be tonnes of articles,answers and explanations on the internet
var percentageDifference = Math.Abs(((double)difference / -1.0d) * 100d);
//Check if the difference is atleast 50%, if so, we assume it as an edge
return(percentageDifference >= 50.0d);
}).Any(b => b);
if (shouldAdd)
{
//Add the pixel as an edge
target.SetPixel(x, y, pixelColor);
}
}
}
}
//We cannot have any messy values in the image's pixels so we convert all of them into white color (-1 is white)
for (var i = 0; i < target.Bits.Length; i++)
{
if (target.Bits[i] == 0)
{
target.Bits[i] = -1;
}
}
//return the edge detected image
return(target.Bitmap);
}
