cyanfish/naps2
1
1
Fork 0
mirror of https://github.com/cyanfish/naps2.git synced 2024-10-27 01:22:49 +03:00
Code Issues Packages Projects Releases Wiki Activity

Move pre-processing inside TesseractOcrEngine

Browse Source
This commit is contained in:
Ben Olden-Cooligan 2024-02-18 10:38:06 -08:00
parent b12d9b53c2
commit 91717d1750
3 changed files with 22 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
NAPS2.Sdk/Ocr/TesseractOcrEngine.cs
View File

@ -70,6 +70,10 @@ public class TesseractOcrEngine : IOcrEngine
string tempHocrFilePathWithExt = tempHocrFilePath + ".hocr";
try
{
if (ocrParams.Mode is OcrMode.FastWithPreProcess or OcrMode.BestWithPreProcess)
{
PreProcessImage(scanningContext, imagePath);
}
var startInfo = new ProcessStartInfo
{
FileName = _tesseractPath,
@ -192,6 +196,21 @@ public class TesseractOcrEngine : IOcrEngine
public event EventHandler? OcrTimeout;
private static void PreProcessImage(ScanningContext scanningContext, string imagePath)
{
IMemoryImage? image = null;
try
{
image = scanningContext.ImageContext.Load(imagePath);
image = image.PerformTransform(new CorrectionTransform(CorrectionMode.Document));
image.Save(imagePath);
}
finally
{
image?.Dispose();
}
}
private static string? GetNearestAncestorAttribute(XElement x, string attributeName)
{
var ancestor = x.AncestorsAndSelf().FirstOrDefault(x => x.Attribute(attributeName) != null);

13
NAPS2.Sdk/Pdf/PdfExporter.cs
View File

@ -344,16 +344,9 @@ public class PdfExporter
// Save the image to a file for use in OCR.
// We don't need to delete this file as long as we pass it to OcrRequestQueue.Enqueue, which takes
// ownership and guarantees its eventual deletion.
if (state.OcrParams!.Mode is OcrMode.FastWithPreProcess or OcrMode.BestWithPreProcess)
{
using var preProcessed = state.Image.WithTransform(new CorrectionTransform(CorrectionMode.Document));
preProcessed.Save(ocrTempFilePath);
}
else
{
using var fileStream = new FileStream(ocrTempFilePath, FileMode.Create, FileAccess.Write);
state.Embedder.CopyToStream(fileStream);
}
// TODO: If the image has transforms and we're saving as JPEG, we should cache to avoid double-rendering
using var fileStream = new FileStream(ocrTempFilePath, FileMode.Create, FileAccess.Write);
state.Embedder.CopyToStream(fileStream);
}
// Start OCR

6
NAPS2.Sdk/Scan/Internal/RemotePostProcessor.cs
View File

@ -1,6 +1,5 @@
using Microsoft.Extensions.Logging;
using NAPS2.Images.Bitwise;
using NAPS2.Ocr;
namespace NAPS2.Scan.Internal;
@ -191,11 +190,6 @@ internal class RemotePostProcessor : IRemotePostProcessor
{
// TODO: If we use tesseract as a library, this is something that that could potentially improve (i.e. not having to save to disk)
// But then again, that doesn't make as much sense on systems (i.e. linux) where tesseract would be provided as an external package
if (options.OcrParams.Mode is OcrMode.FastWithPreProcess or OcrMode.BestWithPreProcess)
{
bitmap = _scanningContext.ImageContext.PerformTransform(bitmap,
new CorrectionTransform(CorrectionMode.Document));
}
return _scanningContext.SaveToTempFile(bitmap);
}
return null;