void ObjButtonGO_Click(object sender, RoutedEventArgs e)
{
int pdf_number = Convert.ToInt32(ObjTextDoc.Text);
int page_number = Convert.ToInt32(ObjTextPage.Text);
string pdf_filename = String.Format(@"C:\temp\{0}.pdf", pdf_number);
Logging.Info("+Rendering page");
MemoryStream ms = MuPDFRenderer.RenderPDFPage(pdf_filename, page_number, 200, null, ProcessPriorityClass.Normal);
BitmapSource bitmap_image = BitmapImageTools.LoadFromBytes(ms.ToArray());
Bitmap bitmap = new Bitmap(ms);
Logging.Info("-Rendering page");
this.Image = bitmap_image;
Logging.Info("+Finding regions");
this.region_locator = new PDFRegionLocator(bitmap);
Logging.Info("-Finding regions");
Recalc();
}
public static WordList DoOCR(string pdf_filename, int page_number)
{
Logging.Info("+Rendering page {1} for PDF file {0}", pdf_filename, page_number);
using (MemoryStream ms = new MemoryStream(SoraxPDFRenderer.GetPageByDPIAsImage(pdf_filename, pdf_user_password, page_number, 200)))
{
Bitmap bitmap = (Bitmap)Image.FromStream(ms);
Logging.Info("-Rendering page #{0}", page_number);
Logging.Info("Startup directory is {0}", Environment.CurrentDirectory);
Logging.Info("Language is '{0}'", language);
using (Tesseract ocr = new Tesseract())
{
ocr.Init(null, language, false);
Logging.Info("+Doing OCR");
const int MIN_WIDTH = 0;
// Build a list of all the rectangles to process
PDFRegionLocator pdf_region_locator = new PDFRegionLocator(bitmap);
PDFRegionLocator.Region last_region = pdf_region_locator.regions[0];
List <Rectangle> rectangles = new List <Rectangle>();
Rectangle last_rectangle = new Rectangle();
foreach (PDFRegionLocator.Region region in pdf_region_locator.regions)
{
int rect_height = region.y - last_region.y;
bool alarming_height = (rect_height <= 0);
Rectangle rectangle = new Rectangle();
if (last_region.state == PDFRegionLocator.SegmentState.BLANKS)
{
// LHS
{
rectangle = new Rectangle(0, last_region.y, bitmap.Width / 2, Math.Max(MIN_WIDTH, rect_height));
}
// RHS
{
rectangle = new Rectangle(bitmap.Width / 2, last_region.y, bitmap.Width / 2, Math.Max(MIN_WIDTH, rect_height));
}
}
else if (last_region.state == PDFRegionLocator.SegmentState.PIXELS)
{
// Full column
{
rectangle = new Rectangle(0, last_region.y, bitmap.Width, Math.Max(MIN_WIDTH, rect_height));
}
}
if (alarming_height || rectangle.Height <= 0)
{
Logging.Warn("Calculated region height is negative or zero: {0} :: Calculated region {1} <-- CURRENT:{2} - LAST:{3}", rect_height, rectangle, region, last_region);
// skip rectangle
}
else if (last_rectangle.X == rectangle.X && last_rectangle.Y == rectangle.Y)
{
Logging.Warn("Overlapping subsequent rectangles will be merged :: CURRENT:{0} - LAST:{1}", rectangle, last_rectangle);
last_rectangle.Width = Math.Max(last_rectangle.Width, rectangle.Width);
last_rectangle.Height = Math.Max(last_rectangle.Height, rectangle.Height);
Logging.Warn("--> Updated 'last' rectangle:{0}", last_rectangle);
}
else
{
rectangles.Add(rectangle);
last_rectangle = rectangle;
}
last_region = region;
}
// DEBUG CODE: Draw in the region rectangles
//
// When we run in NOKILL mode, we "know" we're running in a debugger or stand-alone environment
// intended for testing this code. Hence we should dump the regions image as part of the process.
if (no_kill)
{
string bitmap_diag_path = pdf_filename + @"." + page_number + @"-ocr.png";
Logging.Info("Dumping regions-augmented page {0} PNG image to file {1}", page_number, bitmap_diag_path);
Graphics g = Graphics.FromImage(bitmap);
foreach (Rectangle rectangle in rectangles)
{
if (rectangle.Width <= MIN_WIDTH && rectangle.Height > MIN_WIDTH)
{
DrawRectangleOutline(g, Pens.Purple, rectangle);
}
else if (rectangle.Width > MIN_WIDTH && rectangle.Height <= MIN_WIDTH)
{
DrawRectangleOutline(g, Pens.PowderBlue, rectangle);
}
else if (rectangle.Width <= MIN_WIDTH && rectangle.Height <= MIN_WIDTH)
{
DrawRectangleOutline(g, Pens.Red, rectangle);
}
else
{
DrawRectangleOutline(g, Pens.LawnGreen, rectangle);
}
}
bitmap.Save(bitmap_diag_path, ImageFormat.Png);
}
// Do the OCR on each of the rectangles
WordList word_list = new WordList();
foreach (Rectangle rectangle in rectangles)
{
if (0 == rectangle.Width || 0 == rectangle.Height)
{
Logging.Info("Skipping zero extent rectangle {0}", rectangle.ToString());
continue;
}
Logging.Info("Doing OCR for region {0} on bitmap WxH: {1}x{2}", rectangle.ToString(), bitmap.Width, bitmap.Height);
List <Word> result = ocr.DoOCR(bitmap, rectangle);
Logging.Info("Got {0} words", result.Count);
word_list.AddRange(ConvertToWordList(result, rectangle, bitmap));
}
Logging.Info("-Doing OCR");
Logging.Info("Found {0} words ({1} @ #{2})", word_list.Count, pdf_filename, page_number);
#if false
Logging.Info("+Reordering words for columns");
WordList word_list_ordered = ColumnWordOrderer.ReorderWords(word_list);
Logging.Info("-Reordering words for columns");
word_list_ordered.WriteToFile(ocr_output_filename);
#endif
return(word_list);
}
}
}
public static WordList DoOCR(string pdf_filename, int page_number)
{
Logging.Info("+Rendering page");
SoraxPDFRenderer renderer = new SoraxPDFRenderer(pdf_filename, pdf_user_password, pdf_user_password);
Bitmap bitmap = (Bitmap)Image.FromStream(new MemoryStream(renderer.GetPageByDPIAsImage(page_number, 200)));
Logging.Info("-Rendering page");
Logging.Info("Startup directory is {0}", Environment.CurrentDirectory);
Logging.Info("Language is '{0}'", language);
Tesseract ocr = new Tesseract();
ocr.Init(null, language, false);
Logging.Info("+Doing OCR");
// Build a list of all the rectangles to process
PDFRegionLocator pdf_region_locator = new PDFRegionLocator(bitmap);
PDFRegionLocator.Region last_region = pdf_region_locator.regions[0];
List <Rectangle> rectangles = new List <Rectangle>();
foreach (PDFRegionLocator.Region region in pdf_region_locator.regions)
{
if (false)
{
}
else if (last_region.state == PDFRegionLocator.SegmentState.BLANKS)
{
// LHS
{
Rectangle rectangle = new Rectangle(0, last_region.y, bitmap.Width / 2, region.y - last_region.y);
rectangles.Add(rectangle);
}
// RHS
{
Rectangle rectangle = new Rectangle(bitmap.Width / 2, last_region.y, bitmap.Width / 2, region.y - last_region.y);
rectangles.Add(rectangle);
}
}
else if (last_region.state == PDFRegionLocator.SegmentState.PIXELS)
{
// Full column
{
Rectangle rectangle = new Rectangle(0, last_region.y, bitmap.Width, region.y - last_region.y);
rectangles.Add(rectangle);
}
}
last_region = region;
}
// DEBUG CODE: Draw in the region rectangles
//{
// Graphics g = Graphics.FromImage(bitmap);
// foreach (Rectangle rectangle in rectangles)
// {
// g.DrawRectangle(Pens.Black, rectangle);
// }
// bitmap.Save(@"C:\temp\aaaaaa.png", ImageFormat.Png);
//}
// Do the OCR on each of the rectangles
WordList word_list = new WordList();
foreach (Rectangle rectangle in rectangles)
{
if (0 == rectangle.Width || 0 == rectangle.Height)
{
Logging.Info("Skipping zero extent rectangle {0}", rectangle.ToString());
continue;
}
Logging.Info("Doing OCR for region {0}", rectangle.ToString());
List <Word> result = ocr.DoOCR(bitmap, rectangle);
Logging.Info("Got {0} words", result.Count);
word_list.AddRange(ConvertToWordList(result, rectangle, bitmap));
}
Logging.Info("-Doing OCR");
Logging.Info("Found {0} words", word_list.Count);
//Logging.Info("+Reordering words for columns");
//WordList word_list_ordered = ColumnWordOrderer.ReorderWords(word_list);
//Logging.Info("-Reordering words for columns");
//word_list_ordered.WriteToFile(ocr_output_filename);
return(word_list);
}