feat: add coverage_threshold to skip OCR for small images

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

docling/datamodel/pipeline_options.py

@@ -22,6 +22,9 @@ class TableStructureOptions(BaseModel):
 
 class OcrOptions(BaseModel):
     kind: str
+    coverage_threshold: float = (
+        0.3  # percentage of the area which must be covered by bitmaps for triggering OCR
+    )
 
 
 class EasyOcrOptions(OcrOptions):

docling/models/base_ocr_model.py

@@ -68,6 +68,10 @@ class BaseOcrModel:
             bitmap_rects = []
         coverage, ocr_rects = find_ocr_rects(page.size, bitmap_rects)
 
+        # skip OCR if the bitmap area on the page is smaller than the options threshold
+        if coverage < self.options.coverage_threshold:
+            return []
+
         # return full-page rectangle if sufficiently covered with bitmaps
         if coverage > BITMAP_COVERAGE_TRESHOLD:
             return [

tests/test_options.py

@@ -42,3 +42,23 @@ def test_e2e_conversions(test_doc_path):
         doc_result: ConversionResult = converter.convert(test_doc_path)
 
         assert doc_result.status == ConversionStatus.SUCCESS
+
+
+def test_ocr_coverage_threshold(test_doc_path):
+    pipeline_options = PdfPipelineOptions()
+    pipeline_options.do_ocr = True
+    pipeline_options.ocr_options.coverage_threshold = 1.1
+
+    converter = DocumentConverter(
+        format_options={
+            InputFormat.PDF: PdfFormatOption(
+                pipeline_options=pipeline_options,
+            )
+        }
+    )
+
+    test_doc_path = Path("./tests/data_scanned/ocr_test.pdf")
+    doc_result: ConversionResult = converter.convert(test_doc_path)
+
+    # this should have generated no results, since we set a very high threshold
+    assert len(doc_result.document.texts) == 0