static void Main(string[] args)
{
var metapath = ".\\..\\..\\..\\demo\\4.input.jpg.meta.json";
var meta = Newtonsoft.Json.JsonConvert.DeserializeObject <OCR.Shared.ImageInfo>(File.ReadAllText(metapath));
while (true)
{
var pos = Console.ReadLine().Split(' ').Select(p => int.Parse(p)).ToArray();
var rect = new OCR.Shared.Rect {
X = pos[0], Y = pos[1], W = pos[2], H = pos[3]
};
var filtered = meta.Words.Where(
p =>
p.GravityCenter.X >= rect.X && p.GravityCenter.X <= rect.X + rect.W && p.GravityCenter.Y >= rect.Y && p.GravityCenter.Y <= rect.Y + rect.H ||
p.GeometryCenter.X >= rect.X && p.GeometryCenter.X <= rect.X + rect.W && p.GeometryCenter.Y >= rect.Y && p.GeometryCenter.Y <= rect.Y + rect.H).ToArray();
var newrect = new OCR.Shared.Rect {
X = filtered.Min(p => p.Position.X), Y = filtered.Min(p => p.Position.Y), W = 0, H = 0
};
foreach (var word in filtered)
{
var width = word.Position.X - newrect.X + word.Position.W;
var height = word.Position.Y - newrect.Y + word.Position.H;
if (newrect.W < width)
{
newrect.W = width;
}
if (newrect.H < height)
{
newrect.H = height;
}
}
var bytes = new byte[newrect.W * newrect.H];
for (var i = 0; i < newrect.W * newrect.H; i++)
{
bytes[i] = 0xFF;
}
foreach (var word in filtered)
{
var offsetx = word.Position.X - newrect.X;
var offsety = word.Position.Y - newrect.Y;
for (var j = 0; j < word.Position.H; j++)
{
for (var i = 0; i < word.Position.W; i++)
{
var d = word.Data[j * word.Position.W + i];
if (d == 0xFF)
{
continue;
}
bytes[(offsety + j) * newrect.W + offsetx + i] = d;
}
}
}
var umbmp = UnmanagedImage.FromByteArray(bytes, newrect.W, newrect.H, System.Drawing.Imaging.PixelFormat.Format8bppIndexed);
umbmp.ToManagedImage().Save(meta.FileName + "." + newrect.X + "." + newrect.Y + "." + newrect.W + "." + newrect.H + ".crop.jpg", System.Drawing.Imaging.ImageFormat.Png);
}
}
static void Process2(string fileName)
{
Console.WriteLine(fileName);
var meta = new OCR.Shared.ImageInfo
{
FileName = fileName,
MaskName = fileName + ".mask.png",
FilteredName = fileName + ".filtered.png",
Words = new List <OCR.Shared.Blob>(),
Areas = new List <OCR.Shared.Rect>()
};
using (var img = System.Drawing.Image.FromFile(fileName))
using (var bmp = new Bitmap(img))
{
var gray = new FiltersSequence(Grayscale.CommonAlgorithms.BT709, new GaussianSharpen(), new ContrastCorrection(20))
.Apply(UnmanagedImage.FromManagedImage(bmp));
gray.ToManagedImage().Save(fileName + ".gray.png", System.Drawing.Imaging.ImageFormat.Png);
var bw = new FiltersSequence(new BradleyLocalThresholding {
PixelBrightnessDifferenceLimit = 0.20f
}, new Invert())
.Apply(gray);
bw.ToManagedImage().Save(fileName + ".bw.png", System.Drawing.Imaging.ImageFormat.Png);
var mask = new FiltersSequence(new Dilation(), new BlobsFiltering(5, 5, gray.Width / 2, gray.Height / 2, false))
.Apply(bw);
mask.ToManagedImage().Save(meta.MaskName, System.Drawing.Imaging.ImageFormat.Png);
{
var ccs = new BlobCounter(mask).GetObjectsRectangles();
var sortedw = ccs.Select(p => p.Width).OrderBy(p => p).ToArray();
var sortedh = ccs.Select(p => p.Height).OrderBy(p => p).ToArray();
meta.MedianWordSize = new OCR.Shared.Point {
X = sortedw[sortedw.Length / 2], Y = sortedh[sortedh.Length / 2]
};
Console.WriteLine($"Median: {meta.MedianWordSize.X}, {meta.MedianWordSize.Y}");
}
var filtered = new FiltersSequence(new Invert(), new Intersect(mask), new Invert())
.Apply(gray);
var filteredBmp = filtered.ToManagedImage();
filteredBmp.Save(meta.FilteredName, System.Drawing.Imaging.ImageFormat.Png);
var rgb = new GrayscaleToRGB().Apply(filtered);
mask = new FiltersSequence(new HorizontalRunLengthSmoothing(meta.MedianWordSize.X * 2), new VerticalRunLengthSmoothing(meta.MedianWordSize.Y * 2))
.Apply(mask);
mask.ToManagedImage().Save(fileName + ".area.png", System.Drawing.Imaging.ImageFormat.Png);
foreach (Rectangle rect in new BlobCounter(mask).GetObjectsRectangles())
{
Drawing.FillRectangle(mask, rect, Color.White);
}
mask.ToManagedImage().Save(fileName + ".rect.png", System.Drawing.Imaging.ImageFormat.Png);
foreach (Rectangle rect in new BlobCounter(mask).GetObjectsRectangles())
{
meta.Areas.Add(new OCR.Shared.Rect {
X = rect.X, Y = rect.Y, W = rect.Width, H = rect.Height
});
Drawing.Rectangle(rgb, rect, Color.Red);
}
new Intersect(mask).ApplyInPlace(bw);
foreach (var blob in new BlobCounter(bw).GetObjects(bw, false))
{
var outRect = new OCR.Shared.Rect {
X = blob.Rectangle.X - 1, Y = blob.Rectangle.Y - 1, W = blob.Rectangle.Width + 2, H = blob.Rectangle.Height + 2
};
if (outRect.X < 0)
{
outRect.X = 0; outRect.W--;
}
if (outRect.Y < 0)
{
outRect.Y = 0; outRect.H--;
}
if (outRect.X + outRect.W > bw.Width)
{
outRect.W = bw.Width - outRect.X;
}
if (outRect.Y + outRect.H > bw.Height)
{
outRect.H = bw.Height - outRect.Y;
}
var gravityCenter = new OCR.Shared.Point {
X = (int)blob.CenterOfGravity.X, Y = (int)blob.CenterOfGravity.Y
};
var geometryCenter = new OCR.Shared.Point {
X = blob.Rectangle.X + blob.Rectangle.Width / 2, Y = blob.Rectangle.Y + blob.Rectangle.Height / 2
};
var bytedata = blob.Image.ToByteArray();
var newbytedata = new byte[outRect.W * outRect.H];
for (var j = 0; j < outRect.H - 2; j++)
{
for (var i = 0; i < outRect.W - 2; i++)
{
newbytedata[j * outRect.W + i + 1] = bytedata[j * (outRect.W - 2) + i];
}
}
var blobImg = new FiltersSequence(new Dilation()).Apply(UnmanagedImage.FromByteArray(newbytedata, outRect.W, outRect.H, System.Drawing.Imaging.PixelFormat.Format8bppIndexed));
var area = new Rectangle(outRect.X, outRect.Y, outRect.W, outRect.H);
Drawing.Rectangle(rgb, area, Color.Blue);
Drawing.FillRectangle(rgb, new Rectangle(geometryCenter.X - 2, geometryCenter.Y - 2, 5, 5), Color.Magenta);
Drawing.FillRectangle(rgb, new Rectangle(gravityCenter.X - 2, gravityCenter.Y - 2, 5, 5), Color.Cyan);
var bits = filteredBmp.LockBits(area, System.Drawing.Imaging.ImageLockMode.ReadOnly, System.Drawing.Imaging.PixelFormat.Format8bppIndexed);
meta.Words.Add(
new OCR.Shared.Blob
{
Id = blob.ID,
Data = new FiltersSequence(new Invert(), new Intersect(blobImg), new Invert()).Apply(new UnmanagedImage(bits)).ToByteArray(),
Position = outRect,
GravityCenter = gravityCenter,
GeometryCenter = geometryCenter
});
filteredBmp.UnlockBits(bits);
}
rgb.ToManagedImage().Save(fileName + ".marked.png", System.Drawing.Imaging.ImageFormat.Png);
File.WriteAllText(fileName + ".meta.json", Newtonsoft.Json.JsonConvert.SerializeObject(meta, Newtonsoft.Json.Formatting.Indented));
}
}
