public async Task <string> GetText(Bitmap image)
{
try
{
if (_tesseractEngine.IsDisposed)
{
_tesseractEngine = new TesseractEngine(TessDataPath, "lit", EngineMode.Default);
}
using var img = Pix.LoadFromMemory(ImageToByte(image));
var imageToRecognize = img
.ConvertRGBToGray()
//.BinarizeSauvola(10, 0.35f, false)
//.BinarizeOtsuAdaptiveThreshold(10, 10, 0, 0, 0)
.Deskew()
;
var processTask = new Task <Page>(() => _tesseractEngine.Process(imageToRecognize));
processTask.Start();
RecognizedPage = await processTask;
return(RecognizedPage.GetText());
}
catch (TesseractException ex)
{
throw ex;
}
}
public string GetTextFromImage(byte[] imageData)
{
using Pix img = Pix.LoadFromMemory(imageData);
using Page page = _engine.Process(img);
return page.GetText();
}
public String DoOCR([FromForm] OcrModel request)
{
string name = request.Image.FileName;
var image = request.Image;
var imageStream = new MemoryStream();
if (image.Length > 0)
{
image.CopyTo(imageStream);
}
string tessPath = Path.Combine(trainedDataFolderName, "");
string result = "";
// TODO: Create one instance of engine and inject into app
using (var engine = new TesseractEngine(tessPath, request.DestinationLanguage, EngineMode.Default))
{
// whitelist numbers only
engine.SetVariable("tessedit_char_whitelist", "0123456789");
var img = Pix.LoadFromMemory(imageStream.GetBuffer());
var page = engine.Process(img);
result = page.GetText();
Console.WriteLine(result);
}
return(String.IsNullOrWhiteSpace(result) ? "Ocr is finished. Return empty" : result);
}
public TesseractImage ReadImage(TesseractImage imagem)
{
string path = GetDirectoryName();
byte[] bytesImg = StringHelper.ConvertBase64ToByteArray(imagem.Base64);
try
{
using (var engine = new TesseractEngine(path, imagem.SiglaLinguagem, EngineMode.Default))
{
using (var img = Pix.LoadFromMemory(bytesImg))
{
using (var dataImg = engine.Process(img))
{
imagem.Texto = dataImg.GetText();
imagem.MeanConfidence = dataImg.GetMeanConfidence();
}
}
}
return(imagem);
}
catch (Exception ex)
{
throw new Exception(ex.Message);
}
}
private static string GetOcr(byte[] dosya)
{
using TesseractEngine engine = new("./tessdata", "tur", EngineMode.LstmOnly);
using Pix pixImage = Pix.LoadFromMemory(dosya);
using Page page = engine.Process(pixImage);
return(page.GetText());
}
public string OCRimage(Image ImageToUse, int Zoomlevel, out double TimeTaken, out float Confidence)
{
DateTime Starttime = DateTime.Now;
//load in image
LoadImage(ImageToUse);
//post process the image
ImageToOCR = AccordImageProcessing.AccordImageProc.ImageProcessing(ImageToOCR, Zoomlevel);
//Convert to Tesseract format
byte[] ImgByte = ToByteArray(ImageToOCR, System.Drawing.Imaging.ImageFormat.Bmp);
Pix img = Pix.LoadFromMemory(ImgByte);
// OCR it
Page page = TesseractOCRCore.Process(img);
//get test
string text = page.GetText();
//Get confidence
Confidence = page.GetMeanConfidence();
//Get Time
DateTime EndTime = DateTime.Now;
TimeTaken = (EndTime - Starttime).TotalSeconds;
page.Dispose();
img.Dispose();
return(text);
}
public static string RecognizeText(byte[] imageBytes)
{
string result = null;
try
{
using (var engine = new TesseractEngine(@"./tessdata", "eng", EngineMode.Default))
{
using (Pix pix = Pix.LoadFromMemory(imageBytes))
{
using (var page = engine.Process(pix))
{
string text = page.GetText();
Trace.WriteLine($"Text (GetText): \r\n{text}");
result = $"Mean confidence is {page.GetMeanConfidence()}\r\nText={text.Trim()}";
}
}
}
}
catch (Exception e)
{
Trace.TraceError(e.ToString());
}
return(result);
}
public string FileTessOCR(byte[] data)
{
var pic = Pix.LoadFromMemory(data);
var res = this.ocr.Process(pic);
var result = res.GetText();
return(result.Trim());
}
private string ProcessImage(Image image)
{
using var ocrEngineBest = new TesseractEngine(TessDataPath, "lit", EngineMode.Default);
using var img = Pix.LoadFromMemory(ImageToByte(image));
var imgGray = img.ConvertRGBToGray();
return(ocrEngineBest.Process(imgGray).GetText());
}
private static Pix BitmapToPix(Bitmap bitmap)
{
// Put the bitmap into a memory stream for Pix to load
using MemoryStream stream = new MemoryStream();
bitmap.Save(stream, System.Drawing.Imaging.ImageFormat.Png);
stream.Position = 0;
return(Pix.LoadFromMemory(stream.ToArray()));
}
public static string RecognizeBlocks(byte[] imageBytes)
{
StringBuilder sb = new StringBuilder();
try
{
using (var engine = new TesseractEngine(@"./tessdata", "eng", EngineMode.Default))
{
using (Pix pix = Pix.LoadFromMemory(imageBytes))
{
using (var page = engine.Process(pix))
{
sb.AppendLine("Text (iterator):");
using (ResultIterator iter = page.GetIterator())
{
iter.Begin();
do
{
do
{
do
{
do
{
if (iter.IsAtBeginningOf(PageIteratorLevel.Block))
{
sb.AppendLine("<BLOCK>");
}
sb.Append(iter.GetText(PageIteratorLevel.Word));
sb.Append(" ");
if (iter.IsAtFinalOf(PageIteratorLevel.TextLine, PageIteratorLevel.Word))
{
sb.AppendLine();
}
} while (iter.Next(PageIteratorLevel.TextLine, PageIteratorLevel.Word));
if (iter.IsAtFinalOf(PageIteratorLevel.Para, PageIteratorLevel.TextLine))
{
sb.AppendLine();
}
} while (iter.Next(PageIteratorLevel.Para, PageIteratorLevel.TextLine));
} while (iter.Next(PageIteratorLevel.Block, PageIteratorLevel.Para));
} while (iter.Next(PageIteratorLevel.Block));
}
}
}
}
}
catch (Exception e)
{
Trace.TraceError(e.ToString());
}
return(sb.ToString());
}
/// <summary>
/// 按理说是进来文件出去字符的
/// </summary>
/// <param name="messageOption"></param>
/// <returns></returns>
public async Task <string> ExecuteAsync(Stream file)
{
var stream = new MemoryStream();
var tg_img = Image.FromStream(file);
var bitmap = await ConvertToGray(new Bitmap(tg_img));
bitmap.Save(stream, System.Drawing.Imaging.ImageFormat.Png);
stream.Position = 0;
var texts = new List <string>();
using (var engine = new TesseractEngine(@"/app/tessdata", "chi_sim", EngineMode.Default)) {
using (var img = Pix.LoadFromMemory(stream.ToArray())) {
using (var page = engine.Process(img)) {
var text = page.GetText();
//Console.WriteLine("Mean confidence: {0}", page.GetMeanConfidence());
//Console.WriteLine("Text (GetText): \r\n{0}", text);
//Console.WriteLine("Text (iterator):");
using (var iter = page.GetIterator()) {
iter.Begin();
do
{
do
{
do
{
do
{
//if (iter.IsAtBeginningOf(PageIteratorLevel.Block)) {
// Console.WriteLine("<BLOCK>");
//}
var inner_text = iter.GetText(PageIteratorLevel.Word);
texts.Add(inner_text);
//Console.Write(inner_text);
//Console.Write(" ");
//if (iter.IsAtFinalOf(PageIteratorLevel.TextLine, PageIteratorLevel.Word)) {
// Console.WriteLine();
//}
} while (iter.Next(PageIteratorLevel.TextLine, PageIteratorLevel.Word));
//if (iter.IsAtFinalOf(PageIteratorLevel.Para, PageIteratorLevel.TextLine)) {
// Console.WriteLine();
//}
} while (iter.Next(PageIteratorLevel.Para, PageIteratorLevel.TextLine));
} while (iter.Next(PageIteratorLevel.Block, PageIteratorLevel.Para));
} while (iter.Next(PageIteratorLevel.Block));
}
}
}
}
return(string.Join("", texts));
}
public string GetText(byte[] bytes)
{
using (var img = Pix.LoadFromMemory(bytes))
{
using (var page = _tesseractEngine.Process(img))
{
}
}
return("");
}
private static string GetOcr(byte[] dosya)
{
using (var engine = new TesseractEngine("./tessdata", "tur", EngineMode.TesseractOnly))
{
using (var pixImage = Pix.LoadFromMemory(dosya))
using (var page = engine.Process(pixImage.Scale(3, 3)))
{
return(page.GetText());
}
}
}
private static Pix GetPix(Image <Rgba32> bitmap)
{
byte[] pngBytes;
using (var stream = new MemoryStream())
{
bitmap.SaveAsBmp(stream);
pngBytes = stream.ToArray();
}
return(Pix.LoadFromMemory(pngBytes));
}
/// <summary>
/// read a bitmap image for a number
/// </summary>
/// <param name="image">image to read</param>
/// <returns>number read from the bitmap</returns>
/// <exception cref="NotSupportedException">No number could be read from image</exception>
private int ReadImageText(Bitmap image)
{
ImageConverter converter = new ImageConverter();
byte[] newimage = (byte[])converter.ConvertTo(image, typeof(byte[]));
using var page = engine.Process(Pix.LoadFromMemory(newimage));
string convert = page.GetText();
if (String.IsNullOrEmpty(convert) || String.IsNullOrWhiteSpace(convert))
{
throw new NotSupportedException();
}
return(Convert.ToInt32(convert));
}
public List <Rectangle> GetSubtitleBoundingBoxes(IMagickImage mImage, Rectangle subtitleRegion,
PageIteratorLevel pageIteratorLevel, string cropRegionPath = "")
{
//Preprocess
var imageBytes = PreprocessImage(mImage, subtitleRegion, _preprocessScale, cropRegionPath);
List <Rectangle> boundingBoxes = new List <Rectangle>();
try
{
_tessEngine.SetVariable("user_defined_dpi", 300);
_tessEngine.DefaultPageSegMode = PageSegMode.Auto;
using var img = Pix.LoadFromMemory(imageBytes);
using var page = _tessEngine.Process(img);
//The Iterator way - Same results...
using var iter = page.GetIterator();
iter.Begin();
do
{
if (!iter.TryGetBoundingBox(pageIteratorLevel, out var boundingBoxRect))
{
continue;
}
var boundingBox = new Rectangle(
(int)(boundingBoxRect.X1 / _preprocessScale) + subtitleRegion.X,
(int)(boundingBoxRect.Y1 / _preprocessScale) + subtitleRegion.Y,
(int)(boundingBoxRect.Width / _preprocessScale),
(int)(boundingBoxRect.Height / _preprocessScale));
boundingBoxes.Add(boundingBox);
Console.WriteLine($"Text:{iter.GetText(pageIteratorLevel).Trim()} - Bounding Box:{boundingBox}");
} while (iter.Next(pageIteratorLevel));
return(boundingBoxes);
}
catch (Exception error)
{
Console.WriteLine("Tesseract Error: " + error.Message);
}
Console.WriteLine(boundingBoxes.Count);
return(boundingBoxes);
}
void doOCR()
{
var engine = new TesseractEngine("tessdata", "eng+chi_tra");
Pix tessImg;
using (var ms = new MemoryStream())
{
cropAtRect(img, playArea).Save(ms, System.Drawing.Imaging.ImageFormat.Png);
tessImg = Pix.LoadFromMemory(ms.ToArray());
}
using (var page = engine.Process(tessImg))
{
textBox1.Text = Regex.Replace(page.GetText(), @"\r\n|\r|\n", "\r\n");
}
}
/// <summary>
/// Распознать текст нечитаемого PDF-файла.
/// </summary>
/// <param name="page">Нечитаемая PDF-страница.</param>
/// <returns>Распознанный текст со страницы.</returns>
private static string RecognizePageText(PdfPage page)
{
var options = PdfDrawOptions.Create();
options.BackgroundColor = new PdfRgbColor(255, 255, 255);
options.HorizontalResolution = 200;
options.VerticalResolution = 200;
using var memoryStream = new MemoryStream();
page.Save(memoryStream, options);
using var engine = new TesseractEngine(@"tessdata\fast", "rus+eng", EngineMode.LstmOnly);
using var img = Pix.LoadFromMemory(memoryStream.GetBuffer());
using var recognizedPage = engine.Process(img);
return(recognizedPage.GetText());
}
/// <summary>
/// Main loop for running the scanner
/// </summary>
public static async void MainScanner()
{
ImageConverter converter = new ImageConverter();
while (!runScanner)
{
await Task.Delay(500);
while (runScanner)
{
Bitmap region = CaptureMyScreen(220, 80, 850, 570); //Capture the accept button
using (TesseractEngine engine = new TesseractEngine(@"./tessdata", "eng", EngineMode.Default))
{
using (Pix img = Pix.LoadFromMemory((byte[])converter.ConvertTo(region, typeof(byte[]))))
{
using (Page page = engine.Process(img))
{
string text = page.GetText();
float confidence = page.GetMeanConfidence();
if (text.ToLower().Contains("accept") && confidence > 0.80) //Pretty sure its right
{
System.Windows.Forms.Cursor.Position = new System.Drawing.Point((int)tempxstartpos + 10, (int)tempystartpos + 10);
Thread.Sleep(100);
uint X = (uint)System.Windows.Forms.Cursor.Position.X;
uint Y = (uint)System.Windows.Forms.Cursor.Position.Y;
mouse_event(MOUSEEVENTF_LEFTDOWN | MOUSEEVENTF_LEFTUP, X, Y, 0, 0); //Left click
mouse_event(MOUSEEVENTF_LEFTDOWN | MOUSEEVENTF_LEFTUP, X, Y, 0, 0); //Left click twice for good measure
mouse_event(MOUSEEVENTF_LEFTDOWN | MOUSEEVENTF_LEFTUP, X, Y, 0, 0); //Left click 3 times to be 900% sure!
Environment.Exit(0);
runScanner = false;
}
}
}
}
await Task.Delay(500);
if (!runScanner)
{
//Stopped
}
}
}
}
public static async Task <string> GetParseImageResult(byte[] imageData, string ocrLang)
{
try
{
var langCode = LanguageDictionary.GetLanguageOCRCode(ocrLang);
using (var engine = new TesseractEngine(App.TESSDATA_DICTPATH, langCode, EngineMode.Default))
{
using (var img = Pix.LoadFromMemory(imageData))
{
using (var page = engine.Process(img))
{
var input = page.GetText();
string pattern = @"\n([^A-Z^0-9])";
string substitution = @" $1";
RegexOptions options = RegexOptions.Multiline;
Regex regex = new Regex(pattern, options);
string result = regex.Replace(input, substitution).Trim();
if (result.StartsWith("\n"))
{
result = result.Substring(2);
}
if (result == null || result.Length < 1)
{
throw new Exception("Try again");
}
return(result);
}
}
}
}
catch (Exception e)
{
if (e.Message.Contains("Failed to initialise tesseract engine"))
{
throw new Exception("Can't find data of " + ocrLang);
}
throw e;
}
}
// public string ScanAndOCR()
// {
// var images = WIAScanner.Scan();
// string result = string.Empty;
// using (var ms = new MemoryStream())
// {
// images[0].Save(ms, images[0].RawFormat);
// result = this.api.RecognizeImage(ms);
// }
// return result;
// }
public IEnumerable <string> ScanTessOCR()
{
var images = WIAScanner.Scan();
var result = new List <string>();
using (var ms = new MemoryStream())
{
foreach (var image in images)
{
image.Save(ms, image.RawFormat);
var pic = Pix.LoadFromMemory(ms.ToArray()).Deskew();
pic.Save("ocrImg.png", ImageFormat.Png);
var res = this.ocr.Process(pic);
result.Add(res.GetText().Trim());
}
}
return(result);
}
//Get word location
public string OCRimagewithLocation(Image ImageToUse, int Zoomlevel)
{
//load in image
LoadImage(ImageToUse);
//post process the image
ImageToOCR = AccordImageProcessing.AccordImageProc.ImageProcessing(ImageToOCR, Zoomlevel);
//Convert to Tesseract format
byte[] ImgByte = ToByteArray(ImageToOCR, System.Drawing.Imaging.ImageFormat.Bmp);
Pix img = Pix.LoadFromMemory(ImgByte);
// OCR it
Page page = TesseractOCRCore.Process(img);
//get test
string text = page.GetHOCRText(1);
img.Dispose();
page.Dispose();
return(text);
}
string ImageToText(byte[] imageBytes, RecognitionLanguage rl, RecognitionPrecision rp)
{
try
{
using var engine = new TesseractEngine(@"./Files", _optionService.TranslateLanguage(rl), EngineMode.Default);
using var img = _optionService.TranslatePrecision(rp).Item4 == System.Drawing.Imaging.ImageFormat.Tiff ?
Pix.LoadTiffFromMemory(imageBytes) : Pix.LoadFromMemory(imageBytes);
using var pager = engine.Process(img);
return(pager.GetText().ToString());
//System.Diagnostics.Debug.WriteLine("Mean confidence: {0}", pager.GetMeanConfidence());
//System.Diagnostics.Debug.WriteLine("Text {0}", text);
}
catch (Exception ee)
{
System.Diagnostics.Debug.WriteLine("Unexpected Error: " + ee.Message);
System.Diagnostics.Debug.WriteLine("Details: ");
System.Diagnostics.Debug.WriteLine(ee.ToString());
}
return("");
}
private void Process(MemoryStream ms, PixelFormat pixelFormat, PixelFormat processedPixelFormat, IFileInfo fi, DicomFile dicomFile, string sopID, string studyID, string seriesID, string modality, string[] imageType, int rotationIfAny = 0)
{
float meanConfidence;
string text;
var bytes = ms.ToArray();
// targetBmp is now in the desired format.
using (var page = _tesseractEngine.Process(Pix.LoadFromMemory(bytes)))
{
text = page.GetText();
text = Regex.Replace(text, @"\t|\n|\r", " "); // XXX abrooks surely more useful to have a space?
text = text.Trim();
meanConfidence = page.GetMeanConfidence();
}
//if we find some text
if (!string.IsNullOrWhiteSpace(text))
{
string problemField = rotationIfAny != 0 ? "PixelData" + rotationIfAny : "PixelData";
if (text.Length < _ignoreTextLessThan)
{
_logger.Debug($"Ignoring pixel data discovery in {fi.Name} of length {text.Length} because it is below the threshold {_ignoreTextLessThan}");
}
else
{
var f = factory.Create(fi, dicomFile, text, problemField, new[] { new FailurePart(text, FailureClassification.PixelText) });
AddToReports(f);
_tesseractReport.FoundPixelData(fi, sopID, pixelFormat, processedPixelFormat, studyID, seriesID, modality, imageType, meanConfidence, text.Length, text, rotationIfAny);
}
}
}
public static string ReadFromScreenRect(int x, int y, int width, int height)
{
string text = "";
Bitmap bitmap = Image.Capture(x, y, width, height);
try
{
using (var engine = new TesseractEngine("./tessdata", "eng"))
{
using (var image = Pix.LoadFromMemory(ImageToByte(bitmap)))
{
using (var page = engine.Process(image))
{
text = page.GetText();
}
}
}
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
}
return(text);
}
private static unsafe Pix LoadBitmapToPix(Bitmap inBitmap)
{
Pix pix = null;
string uniqId = Guid.NewGuid().ToString("D");
// Clone the bitmap from whatever format it's in now to a 24bit bitmap
using (var bitmap = inBitmap.Clone(new Rectangle(0, 0, inBitmap.Width, inBitmap.Height), PixelFormat.Format24bppRgb))
{
using (MemoryStream stream = new MemoryStream())
{
bitmap.Save(stream, System.Drawing.Imaging.ImageFormat.Png);
stream.Position = 0;
// Write out the result for later processing
using (FileStream file = File.OpenWrite($"ProcessedImages/{uniqId}_Original.png"))
{
stream.CopyTo(file);
stream.Position = 0;
}
}
BitmapData bmpData = bitmap.LockBits(new Rectangle(0, 0, bitmap.Width, bitmap.Height), ImageLockMode.ReadWrite, bitmap.PixelFormat);
byte bpp = 24;
byte * scan0 = (byte *)bmpData.Scan0.ToPointer();
int[] scales = new int[255];
for (int i = 0; i < 255; i++)
{
scales[i] = 0;
}
for (int row = 0; row < bmpData.Height; row++)
{
for (int col = 0; col < bmpData.Width; col++)
{
byte *data = scan0 + row * bmpData.Stride + col * bpp / 8;
byte blue = data[0];
byte green = data[1];
byte red = data[2];
// Calculate grayscale and invert
byte avg = (byte)(0.2989 * red + 0.5870 * green + 0.1140 * blue);
blue = green = red = (byte)(255 - avg);
// Generate a curve for finding the best spot to threshold
scales[avg]++;
data[0] = blue;
data[1] = green;
data[2] = red;
}
}
int max = 0;
byte maxColor = 0;
for (byte i = 0; i < 255; i++)
{
if (scales[i] > max)
{
max = scales[i];
maxColor = i;
}
}
// Take 100 off the color, and use that
byte threshold = (byte)(255 - maxColor - 130);
// Apply the threshold to all the bits
for (int row = 0; row < bmpData.Height; row++)
{
for (int col = 0; col < bmpData.Width; col++)
{
byte *data = scan0 + row * bmpData.Stride + col * bpp / 8;
byte blue = data[0];
byte green = data[1];
byte red = data[2];
// Calculate grayscale and invert
byte pixel = (byte)(red > threshold ? 255 : 0);
blue = green = red = pixel;
data[0] = blue;
data[1] = green;
data[2] = red;
}
}
bitmap.UnlockBits(bmpData);
// Put the bitmap into a memory stream for Pix to load
using (MemoryStream stream = new MemoryStream())
{
bitmap.Save(stream, System.Drawing.Imaging.ImageFormat.Png);
stream.Position = 0;
// Write out the result for later processing
using (FileStream file = File.OpenWrite($"ProcessedImages/{uniqId}_Threshold.png"))
{
stream.CopyTo(file);
stream.Position = 0;
}
pix = Pix.LoadFromMemory(stream.ToArray());
}
}
return(pix);
}
public static Pix ConvertBitmapToPix(Bitmap img)
{
return(Pix.LoadFromMemory(ImageToByteArray(img)));
}
private async Task AnalyseUsingTesseract(StorageFile imageFile)
{
try
{
byte[] result;
using (Stream stream = await imageFile.OpenStreamForReadAsync())
{
using (var memoryStream = new MemoryStream())
{
stream.CopyTo(memoryStream);
result = memoryStream.ToArray();
}
}
string language = "eng";
if (LanguageSelection.SelectedIndex == 1)
{
language = "chi_sim";
}
if (LanguageSelection.SelectedIndex == 2)
{
language = "kor";
}
using (var engine = new TesseractEngine(@"./TrainedData", language, EngineMode.Default))
{
using (var img = Pix.LoadFromMemory(result))
{
using (var page = engine.Process(img))
{
string text = "=======================\r\nPLAIN TEXT\r\n=======================\r\n";
text += page.GetText();
OutputTextTitle.Text = string.Format("Output Text (Mean Confidence: {0})", page.GetMeanConfidence());
var recognisedTexts = new List <RecognisedText>();
var pgLevel = PageIteratorLevel.Word;
using (var iter = page.GetIterator())
{
do
{
if (iter.TryGetBoundingBox(pgLevel, out Rect boundary))
{
var reading = iter.GetText(pgLevel);
recognisedTexts.Add(new RecognisedText
{
Text = reading,
Boundary = boundary
});
}
} while (iter.Next(pgLevel));
}
text += "=======================\r\nJSON REPRESENTATION\r\n=======================\r\n";
text += JsonSerializer.Serialize(recognisedTexts);
OutputText.Document.SetText(TextSetOptions.FormatRtf, text);
}
}
}
}
catch (System.Reflection.TargetInvocationException ex)
{
await AnalyseUsingAzureComputerVisionAsync(imageFile);
}
catch (Exception ex)
{
await ShowExceptionMessageAsync(ex);
}
}
