public static void CreateXmpMetadata(PdfDocument document, PdfCompat compat)
{
var metadataDict = new PdfDictionary(document);
metadataDict.Elements["/Type"] = new PdfName("/Metadata");
metadataDict.Elements["/Subtype"] = new PdfName("/XML");
metadataDict.CreateStream(CreateRawXmpMetadata(document.Info, GetConformance(compat)));
document.Internals.AddObject(metadataDict);
document.Internals.Catalog.Elements["/Metadata"] = metadataDict.Reference;
}
private static void DrawImageOnPage(PdfPage page, XImage img, PdfCompat compat)
{
if (compat != PdfCompat.Default)
{
img.Interpolate = false;
}
Size realSize = GetRealSize(img);
page.Width = realSize.Width;
page.Height = realSize.Height;
using (XGraphics gfx = XGraphics.FromPdfPage(page))
{
gfx.DrawImage(img, 0, 0, realSize.Width, realSize.Height);
}
}
private static (string, string) GetConformance(PdfCompat compat)
{
switch (compat)
{
case PdfCompat.PdfA1B:
return("1", "B");
case PdfCompat.PdfA2B:
return("2", "B");
case PdfCompat.PdfA3B:
return("3", "B");
case PdfCompat.PdfA3U:
return("3", "U");
default:
return("", "");
}
}
private bool BuildDocumentWithOcr(Func<int, bool> progressCallback, PdfDocument document, PdfCompat compat, IEnumerable<ScannedImage> images, string ocrLanguageCode)
{
// Use a pipeline so that multiple pages/images can be processed in parallel
// Note: No locks needed on the document because the design of the pipeline ensures no two threads will work on it at once
int progress = 0;
Pipeline.For(images).Step(image =>
{
// Step 1: Load the image into memory, draw it on a new PDF page, and save a copy of the processed image to disk for OCR
if (!progressCallback(progress))
{
return null;
}
bool importedPdfPassThrough = image.FileFormat == null && !image.RecoveryIndexImage.TransformList.Any();
PdfPage page;
if (importedPdfPassThrough)
{
page = CopyPdfPageToDoc(document, image);
// Scan through the page looking for text
var elements = page.Contents.Elements;
for (int i = 0; i < elements.Count; i++)
{
string textAndFormatting = elements.GetDictionary(i).Stream.ToString();
var reader = new StringReader(textAndFormatting);
bool inTextBlock = false;
string line;
while ((line = reader.ReadLine()) != null)
{
if (line.EndsWith("BT"))
{
inTextBlock = true;
}
else if (line.EndsWith("ET"))
{
inTextBlock = false;
}
else if (inTextBlock &&
(line.EndsWith("TJ") || line.EndsWith("Tj")
|| line.EndsWith("\"") || line.EndsWith("'")))
{
// Text-showing operators
// Since this page already contains text, don't use OCR
return null;
}
}
}
}
else
{
page = document.AddPage();
}
using (Stream stream = scannedImageRenderer.RenderToStream(image))
using (var img = XImage.FromStream(stream))
{
if (!progressCallback(progress))
{
return null;
}
if (!importedPdfPassThrough)
{
DrawImageOnPage(page, img, compat);
}
if (!progressCallback(progress))
{
return null;
}
string tempImageFilePath = Path.Combine(Paths.Temp, Path.GetRandomFileName());
img.GdiImage.Save(tempImageFilePath);
return Tuple.Create(page, tempImageFilePath);
}
}).StepParallel((page, tempImageFilePath) =>
{
// Step 2: Run OCR on the processsed image file
// This step is doubly parallel since not only can it run alongside other stages of the pipeline,
// multiple files can also be OCR'd at once (no interdependencies, it doesn't touch the document)
OcrResult ocrResult;
try
{
if (!progressCallback(progress))
{
return null;
}
// ReSharper disable once AccessToModifiedClosure
ocrResult = ocrEngine.ProcessImage(tempImageFilePath, ocrLanguageCode, () => !progressCallback(progress));
}
finally
{
File.Delete(tempImageFilePath);
}
// The final pipeline step is pretty fast, so updating progress here is more accurate
if (progressCallback(progress))
{
Interlocked.Increment(ref progress);
progressCallback(progress);
}
return Tuple.Create(page, ocrResult);
}).StepBlock().Run((page, ocrResult) =>
{
// Step 3: Draw the OCR text on the PDF page
if (ocrResult == null)
{
return;
}
if (!progressCallback(progress))
{
return;
}
DrawOcrTextOnPage(page, ocrResult);
});
return progressCallback(progress);
}
private bool BuildDocumentWithoutOcr(Func<int, bool> progressCallback, PdfDocument document, PdfCompat compat, IEnumerable<ScannedImage> images)
{
int progress = 0;
foreach (var image in images)
{
bool importedPdfPassThrough = image.FileFormat == null && !image.RecoveryIndexImage.TransformList.Any();
if (importedPdfPassThrough)
{
CopyPdfPageToDoc(document, image);
}
else
{
using (Stream stream = scannedImageRenderer.RenderToStream(image))
using (var img = XImage.FromStream(stream))
{
if (!progressCallback(progress))
{
return false;
}
PdfPage page = document.AddPage();
DrawImageOnPage(page, img, compat);
}
}
progress++;
}
return true;
}
private bool BuildDocumentWithOcr(ProgressHandler progressCallback, CancellationToken cancelToken, PdfDocument document, PdfCompat compat, ICollection <ScannedImage.Snapshot> snapshots, IOcrEngine ocrEngine, OcrParams ocrParams)
{
int progress = 0;
progressCallback(progress, snapshots.Count);
List <(PdfPage, Task <OcrResult>)> ocrPairs = new List <(PdfPage, Task <OcrResult>)>();
// Step 1: Create the pages, draw the images, and start OCR
foreach (var snapshot in snapshots)
{
if (cancelToken.IsCancellationRequested)
{
break;
}
bool importedPdfPassThrough = snapshot.Source.FileFormat == null && !snapshot.TransformList.Any();
PdfPage page;
if (importedPdfPassThrough)
{
page = CopyPdfPageToDoc(document, snapshot.Source);
if (PageContainsText(page))
{
// Since this page already contains text, don't use OCR
continue;
}
}
else
{
page = document.AddPage();
}
string tempImageFilePath = Path.Combine(Paths.Temp, Path.GetRandomFileName());
using (Stream stream = scannedImageRenderer.RenderToStream(snapshot).Result)
using (var img = XImage.FromStream(stream))
{
if (cancelToken.IsCancellationRequested)
{
break;
}
if (!importedPdfPassThrough)
{
DrawImageOnPage(page, img, compat);
}
if (cancelToken.IsCancellationRequested)
{
break;
}
if (!ocrRequestQueue.HasCachedResult(ocrEngine, snapshot, ocrParams))
{
img.GdiImage.Save(tempImageFilePath);
}
}
if (cancelToken.IsCancellationRequested)
{
File.Delete(tempImageFilePath);
break;
}
// Start OCR
var ocrTask = ocrRequestQueue.QueueForeground(ocrEngine, snapshot, tempImageFilePath, ocrParams, cancelToken);
ocrTask.ContinueWith(task =>
{
// This is the best place to put progress reporting
// Long-running OCR is done, and drawing text on the page (step 2) is very fast
if (!cancelToken.IsCancellationRequested)
{
Interlocked.Increment(ref progress);
progressCallback(progress, snapshots.Count);
}
}, TaskContinuationOptions.ExecuteSynchronously);
// Record the page and task for step 2
ocrPairs.Add((page, ocrTask));
}
// Step 2: Wait for all the OCR results, and draw the text on each page
foreach (var(page, ocrTask) in ocrPairs)
{
if (cancelToken.IsCancellationRequested)
{
break;
}
if (ocrTask.Result == null)
{
continue;
}
DrawOcrTextOnPage(page, ocrTask.Result);
}
return(!cancelToken.IsCancellationRequested);
}
private bool BuildDocumentWithoutOcr(ProgressHandler progressCallback, CancellationToken cancelToken, PdfDocument document, PdfCompat compat, ICollection <ScannedImage.Snapshot> snapshots)
{
int progress = 0;
progressCallback(progress, snapshots.Count);
foreach (var snapshot in snapshots)
{
bool importedPdfPassThrough = snapshot.Source.FileFormat == null && !snapshot.TransformList.Any();
if (importedPdfPassThrough)
{
CopyPdfPageToDoc(document, snapshot.Source);
}
else
{
using (Stream stream = scannedImageRenderer.RenderToStream(snapshot).Result)
using (var img = XImage.FromStream(stream))
{
if (cancelToken.IsCancellationRequested)
{
return(false);
}
PdfPage page = document.AddPage();
DrawImageOnPage(page, img, compat);
}
}
progress++;
progressCallback(progress, snapshots.Count);
}
return(true);
}
