public void ShouldThrowExceptionWhenThresholdIsNegative()
{
using (var image = new MagickImage())
{
Assert.Throws <ArgumentException>("settings", () => image.Deskew(new Percentage(-1)));
}
}
public void ShouldThrowExceptionWhenSettingsIsNull()
{
using (var image = new MagickImage())
{
Assert.Throws <ArgumentNullException>("settings", () => image.Deskew(null));
}
}
public override MagickImage ApplyFilter(MagickImage image)
{
image.BackgroundColor = MagickColor.FromRgb(0, 0, 0);
image.Deskew(new Percentage(Threshold));
image.RePage();
return(image);
}
public IEnumerable <byte> PreProcess(IEnumerable <byte> bitmap)
{
using (var processed = new MagickImage(bitmap.ToArray()))
{
processed.Deskew(new Percentage(60));
return(processed.ToByteArray());
}
}
public void ShouldReturnTheAngle()
{
using (var image = new MagickImage(Files.LetterJPG))
{
var angle = image.Deskew(new Percentage(10));
Assert.InRange(angle, 7.01, 7.02);
}
}
public void ShouldReturnTheAngle()
{
using (var image = new MagickImage(Files.LetterJPG))
{
var angle = image.Deskew(new Percentage(10));
Assert.AreEqual(7.01, angle, 0.01);
}
}
private void UnrotateImage(MagickImage image)
{
if (!Unrotate)
{
return;
}
image.BackgroundColor = BackgroundColor;
image.Deskew((Percentage)40);
}
public void CreateNewImageFrom(string source, string target)
{
using (MagickImage image = new MagickImage(source))
{
image.Deskew(DeskewPercentage);
image.Despeckle();
image.Grayscale(PixelIntensityMethod.Rec709Luminance);
image.Write(target);
}
}
public void ShouldThrowExceptionWhenSettingsThresholdIsNegative()
{
using (var image = new MagickImage())
{
var settings = new DeskewSettings
{
Threshold = new Percentage(-1),
};
Assert.Throws <ArgumentException>("settings", () => image.Deskew(settings));
}
}
public void ShouldDeskewTheImage()
{
using (var image = new MagickImage(Files.LetterJPG))
{
image.ColorType = ColorType.Bilevel;
ColorAssert.Equal(MagickColors.White, image, 471, 92);
image.Deskew(new Percentage(10));
ColorAssert.Equal(new MagickColor("#007400740074"), image, 471, 92);
}
}
public void ShouldUseAutoCrop()
{
using (var image = new MagickImage(Files.LetterJPG))
{
var settings = new DeskewSettings
{
AutoCrop = true,
Threshold = new Percentage(10),
};
image.Deskew(settings);
Assert.Equal(480, image.Width);
Assert.Equal(577, image.Height);
}
}
private void bwImageProcess_DoWork(object sender, DoWorkEventArgs e)
{
if (this.m_processQueue.Count > 0)
{
string fname = "";
lock (this.m_syncObject)
{
fname = m_processQueue.Dequeue();
}
byte[] tempImg = File.ReadAllBytes(fname);
string ocrText = "";
MagickReadSettings settings = new MagickReadSettings();
using (MagickImage image = new MagickImage(tempImg, settings))
{
image.Quality = scannedQuality;
image.Deskew(new Percentage(100));
image.Trim();
image.AutoOrient();
File.Delete(fname);
image.Write(fname);
Bitmap bmp = Grayscale.CommonAlgorithms.BT709.Apply(image.ToBitmap());
Threshold thresholdFilter = new Threshold(127);
Bitmap searchOcr = thresholdFilter.Apply(bmp);
TesseractEngine engine = new TesseractEngine("tessdata", "eng", EngineMode.Default);
Page page = engine.Process(searchOcr);
ocrText = page.GetText();
}
fileList.Add(fname);
DataRow r = dtDoc.NewRow();
r["DocFilename"] = fname;
r["DocValue"] = "0";
r["DocSize"] = string.Format("{0:n0} KB", Math.Round((double)new FileInfo(fname).Length / 1024));
r["DocOcr"] = ocrText;
dtDoc.Rows.Add(r);
e.Result = r;
}
}
public IEnumerable <byte> PreProcess(IEnumerable <byte> bitmap)
{
using (var processed = new MagickImage(bitmap.ToArray()))
{
processed.ColorSpace = ColorSpace.Gray;
processed.ColorType = ColorType.Grayscale;
processed.Normalize();
// Denoise
using (var denoise = processed.Clone())
{
denoise.ColorSpace = ColorSpace.Gray;
denoise.Negate();
denoise.AdaptiveThreshold(10, 10, new Percentage(5));
denoise.ContrastStretch(new Percentage(0));
processed.Composite(denoise, CompositeOperator.CopyAlpha);
}
processed.Opaque(MagickColors.Transparent, MagickColors.White);
processed.Alpha(AlphaOption.Off);
// Deskew
processed.BackgroundColor = MagickColors.White;
processed.Deskew(new Percentage(40));
// Sharpen
processed.Sharpen(0, 1);
// Saturate
processed.Modulate(new Percentage(100), new Percentage(200), new Percentage(100));
// Trim
processed.Trim();
processed.RePage();
processed.Compose = CompositeOperator.Over;
processed.BorderColor = MagickColors.White;
processed.Border(15);
return(processed.ToByteArray());
}
}
public static void Run([BlobTrigger("inbox/{inputFilename}", Connection = "StorageAccountConnString")] Stream inputStream, string inputFilename,
[Blob("outbox/out-{inputFilename}", FileAccess.Write, Connection = "StorageAccountConnString")] Stream outputStream, ILogger log)
{
log.LogInformation($"STEP: trigger fired for file:{inputFilename}, Size: {inputStream.Length} Bytes");
using (MagickImage image = new MagickImage(inputStream))
{
log.LogInformation($"STEP: Read image {inputFilename} with {image.BaseWidth} x {image.BaseHeight} in {image.Format.ToString()}");
// Step 0 - does nothing in the sample images I used
image.AutoOrient();
// Step 1 - Deskew
image.BackgroundColor = MagickColor.FromRgb(0, 0, 0);
image.Deskew(new Percentage(1)); // documentation suggests "A threshold of 40% works for most images" but 1% seems to work well
IMagickImage rotatedImage = image.Clone();
log.LogInformation($"STEP: Deskewed {inputFilename}");
// Step 2 - Apply threshold to transform in black and white image
image.AutoThreshold(AutoThresholdMethod.OTSU);
log.LogInformation($"STEP: Applied OTSU to {inputFilename}");
// Step 3 - find the regions (blocs) in the image and group them if large enough
IEnumerable <ConnectedComponent> components = null;
ConnectedComponentsSettings ccs = new ConnectedComponentsSettings();
ccs.AreaThreshold = 500 * 500.0; // 500x500 -- seems to be pointless, many more regions are returned
ccs.Connectivity = 8;
components = image.ConnectedComponents(ccs);
// if there are multiple blocs, consolidate them in a larger block
if (components != null && components.Count() > 0)
{
log.LogInformation($"STEP: Looked for regions in {inputFilename}, there are {components.Count()}");
// filter out the smaller rectangles, as the AreaThreshold parameter seems not to be working
List <ConnectedComponent> biggerComponents = components.Where(cc => cc.Height * cc.Width >= 250000 && cc.Height * cc.Width != image.Width * image.Height) /*.OrderByDescending(i => i.Height * i.Width)*/.ToList();
int topLeftX = biggerComponents[0].X, topLeftY = biggerComponents[0].Y, bottomRightX = biggerComponents[0].Width + topLeftX, bottomRightY = biggerComponents[0].Height + topLeftY;
foreach (ConnectedComponent cc in biggerComponents)
{
#region Debug -- draw the regions on the image
//DrawableStrokeColor strokeColor = new DrawableStrokeColor(new MagickColor("yellow"));
//DrawableStrokeWidth stokeWidth = new DrawableStrokeWidth(3);
//DrawableFillColor fillColor = new DrawableFillColor(new MagickColor(50, 50, 50, 128));
//DrawableRectangle dr = new DrawableRectangle(cc.X, cc.Y, cc.X + cc.Width, cc.Y + cc.Height);
//rotatedImage.Draw(dr, strokeColor, stokeWidth, fillColor);
#endregion
if (cc.X < topLeftX)
{
topLeftX = cc.X;
}
if (cc.Y < topLeftY)
{
topLeftY = cc.Y;
}
if (cc.X + cc.Width > bottomRightX)
{
bottomRightX = cc.X + cc.Width;
}
if (cc.Y + cc.Height > bottomRightY)
{
bottomRightY = cc.Y + cc.Height;
}
}
#region Debug -- draw the bounding box on the image
//DrawableStrokeColor strokeColor2 = new DrawableStrokeColor(new MagickColor("purple"));
//DrawableStrokeWidth stokeWidth2 = new DrawableStrokeWidth(3);
//DrawableFillColor fillColor2 = new DrawableFillColor(new MagickColor(50, 50, 50, 128));
//DrawableRectangle dr2 = new DrawableRectangle(topLeftX, topLeftY, bottomRightX, bottomRightY);
//rotatedImage.Draw(dr2, strokeColor2, stokeWidth2, fillColor2);
#endregion
// Step 4 - Crop the image
MagickGeometry mg = new MagickGeometry(topLeftX, topLeftY, bottomRightX - topLeftX, bottomRightY - topLeftY);
rotatedImage.RePage(); // this is needed because otherwise the crop is relative to the page information and sometimes this leads to an incorrect crop
rotatedImage.Crop(mg);
log.LogInformation($"STEP: Cropped {inputFilename} to fit existing large regions");
}
else
{
log.LogInformation($"STEP: Looked for large regions in {inputFilename}, none were found, skipping crop");
}
// Step 5 - Resize the image to 1200px width (todo: move to configuration)
int originalWidth = rotatedImage.BaseWidth;
int originalHeight = rotatedImage.BaseHeight;
rotatedImage.Resize(1200, 0); // make width 1200, height proportional
log.LogInformation($"STEP: Resized {inputFilename} from {originalWidth}x{originalHeight} to {image.BaseWidth}x{image.BaseHeight}");
// Step 6 - write out as Jpeg with 70% quality
rotatedImage.Format = MagickFormat.Jpeg;
rotatedImage.Quality = 70;
rotatedImage.Write(outputStream);
log.LogInformation($"STEP: Wrote out {inputFilename} as JPEG");
}
log.LogInformation($"STEP: Processing of {inputFilename} done");
}
