feat: Enable antialiasing for PDF image extraction to improve OCR accuracy (#7562)

### What problem does this PR solve?

When the PDF uses vector fonts, the rendered text in the captured page
image often has missing strokes, leading to numerous OCR errors and
incorrect characters. Similar issues also occur in the extracted chart
images.

**Before**

![0089e1f76205b5b3](https://github.com/user-attachments/assets/a84f8cd7-48ae-4da4-81ca-fc0bd93320f1)

**After**

![03053149e919773a](https://github.com/user-attachments/assets/45fa5ebb-a2de-42b1-9535-1ea087877eb2)

You can use the following document for testing.

[Casio说明书.pdf](https://github.com/user-attachments/files/20119690/Casio.pdf)


### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
- [ ] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [ ] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):

Co-authored-by: liuzhenghua-jk <liuzhenghua-jk@360shuke.com>
This commit is contained in:
liuzhenghua
2025-05-12 09:50:21 +08:00
committed by GitHub
parent 473aa28422
commit ea5e8caa69

View File

@ -1015,7 +1015,7 @@ class RAGFlowPdfParser:
with sys.modules[LOCK_KEY_pdfplumber]:
with (pdfplumber.open(fnm) if isinstance(fnm, str) else pdfplumber.open(BytesIO(fnm))) as pdf:
self.pdf = pdf
self.page_images = [p.to_image(resolution=72 * zoomin).annotated for i, p in
self.page_images = [p.to_image(resolution=72 * zoomin, antialias=True).annotated for i, p in
enumerate(self.pdf.pages[page_from:page_to])]
try: