Merge pull request #54 from bertsky/set-dpi

Inform Tesseract of true DPI as detected from source image in all processors (notably improves layout recognition). Also fixes #55

ocrd_tesserocr/crop.py

@@ -10,20 +10,22 @@
 from ocrd_models.ocrd_page import (
     MetadataItemType,
     LabelsType, LabelType,
-    CoordsType,
+    CoordsType, AlternativeImageType,
     to_xml
 )
 from ocrd_models.ocrd_page_generateds import BorderType
+from ocrd_models import OcrdExif
 from ocrd import Processor
 
 from .config import TESSDATA_PREFIX, OCRD_TOOL
 from .common import (
     bbox_from_points, points_from_bbox,
-    bbox_from_xywh
+    bbox_from_xywh, save_image_file
 )
 
 TOOL = 'ocrd-tesserocr-crop'
 LOG = getLogger('processor.TesserocrCrop')
+FILEGRP_IMG = 'OCR-D-IMG-CROP'
 
 class TesserocrCrop(Processor):
 
@@ -87,8 +89,15 @@ def process(self):
                                 min_x, max_x, min_y, max_y)
 
                 page_image = self.workspace.resolve_image_as_pil(page.imageFilename)
-                dpi = page_image.info.get('dpi', (300,300))[0]
-                zoom = 300 / dpi
+                page_image_info = OcrdExif(page_image)
+                if page_image_info.xResolution != 1:
+                    dpi = page_image_info.xResolution
+                    if page_image_info.resolutionUnit == 'cm':
+                        dpi = round(dpi * 2.54)
+                    tessapi.SetVariable('user_defined_dpi', str(dpi))
+                    zoom = 300 / dpi
+                else:
+                    zoom = 1
                 LOG.debug("Cropping with tesseract")
                 tessapi.SetImage(page_image)
                 # PSM.SPARSE_TEXT: get as much text as possible in no particular order
@@ -146,7 +155,21 @@ def process(self):
                     LOG.debug("Padded page border: %i:%i,%i:%i", min_x, max_x, min_y, max_y)
                     border = BorderType(Coords=CoordsType(
                         points_from_bbox(min_x, min_y, max_x, max_y)))
+                    # update PAGE (annotate border):
                     page.set_Border(border)
+                    # update METS (add the image file):
+                    page_image = page_image.crop(
+                        box=(min_x, min_y, max_x, max_y))
+                    file_id = input_file.ID.replace(self.input_file_grp, FILEGRP_IMG)
+                    if file_id == input_file.ID:
+                        file_id = concat_padded(FILEGRP_IMG, n)
+                    file_path = save_image_file(self.workspace, page_image,
+                                                file_id,
+                                                page_id=page_id,
+                                                file_grp=FILEGRP_IMG)
+                    # update PAGE (reference the image file):
+                    page.add_AlternativeImage(AlternativeImageType(
+                        filename=file_path, comments="cropped"))
                 else:
                     LOG.error("Cannot find valid extent for page '%s'", page_id)
 

ocrd_tesserocr/deskew.py

@@ -22,6 +22,7 @@
     TextRegionType, PageType,
     to_xml
 )
+from ocrd_models import OcrdExif
 from ocrd import Processor
 
 from .config import TESSDATA_PREFIX, OCRD_TOOL
@@ -85,6 +86,12 @@ def process(self):
                                                                for name in self.parameter.keys()])]))
                 page = pcgts.get_Page()
                 page_image = self.workspace.resolve_image_as_pil(page.imageFilename)
+                page_image_info = OcrdExif(page_image)
+                if page_image_info.xResolution != 1:
+                    dpi = page_image_info.xResolution
+                    if page_image_info.resolutionUnit == 'cm':
+                        dpi = round(dpi * 2.54)
+                    tessapi.SetVariable('user_defined_dpi', str(dpi))
                 LOG.info("Deskewing on '%s' level in page '%s'", oplevel, page_id)
 
                 page_image, page_xywh = image_from_page(
@@ -118,7 +125,11 @@ def process(self):
                     content=to_xml(pcgts))
 
     def _process_segment(self, tessapi, segment, image, xywh, where, page_id, file_id):
-        comments = 'cropped'
+        if (isinstance(segment, PageType) and
+            not xywh['x'] and not xywh['y']):
+            comments = ''
+        else:
+            comments = 'cropped'
         angle = 0.
         tessapi.SetImage(image)
         #tessapi.SetPageSegMode(PSM.AUTO_OSD)
@@ -127,7 +138,7 @@ def _process_segment(self, tessapi, segment, image, xywh, where, page_id, file_i
         #
         osr = tessapi.DetectOrientationScript()
         if osr:
-            assert osr['orient_conf'] and not math.isnan(osr['orient_conf']), \
+            assert not math.isnan(osr['orient_conf']), \
                 "orientation detection failed (Tesseract probably compiled without legacy OEM, or osd model not installed)"
             if osr['orient_conf'] < 10:
                 LOG.info('ignoring OSD orientation result %d° due to low confidence %.0f in %s',
@@ -138,7 +149,7 @@ def _process_segment(self, tessapi, segment, image, xywh, where, page_id, file_i
                 angle = osr['orient_deg']
                 if angle:
                     comments += ',rotated-%d' % angle
-            assert osr['script_conf'] and not math.isnan(osr['script_conf']), \
+            assert not math.isnan(osr['script_conf']), \
                 "script detection failed (Tesseract probably compiled without legacy OEM, or osd model not installed)"
             if osr['script_conf'] < 10:
                 LOG.info('ignoring OSD script result "%s" due to low confidence %.0f in %s',

ocrd_tesserocr/recognize.py

@@ -18,6 +18,7 @@
     TextEquivType, TextStyleType,
     to_xml)
 from ocrd_modelfactory import page_from_file
+from ocrd_models import OcrdExif
 from ocrd import Processor
 
 from .config import TESSDATA_PREFIX, OCRD_TOOL
@@ -133,6 +134,12 @@ def process(self):
                                                                for name in self.parameter.keys()])]))
                 page = pcgts.get_Page()
                 page_image = self.workspace.resolve_image_as_pil(page.imageFilename)
+                page_image_info = OcrdExif(page_image)
+                if page_image_info.xResolution != 1:
+                    dpi = page_image_info.xResolution
+                    if page_image_info.resolutionUnit == 'cm':
+                        dpi = round(dpi * 2.54)
+                    tessapi.SetVariable('user_defined_dpi', str(dpi))
                 page_image, page_xywh = image_from_page(
                     self.workspace, page, page_image, page_id)
                 #tessapi.SetImage(page_image)

ocrd_tesserocr/segment_line.py

@@ -17,6 +17,7 @@
     TextLineType,
     to_xml
 )
+from ocrd_models import OcrdExif
 
 from .config import TESSDATA_PREFIX, OCRD_TOOL
 from .common import (
@@ -71,6 +72,12 @@ def process(self):
                                                                for name in self.parameter.keys()])]))
                 page = pcgts.get_Page()
                 page_image = self.workspace.resolve_image_as_pil(page.imageFilename)
+                page_image_info = OcrdExif(page_image)
+                if page_image_info.xResolution != 1:
+                    dpi = page_image_info.xResolution
+                    if page_image_info.resolutionUnit == 'cm':
+                        dpi = round(dpi * 2.54)
+                    tessapi.SetVariable('user_defined_dpi', str(dpi))
                 page_image, page_xywh = image_from_page(
                     self.workspace, page, page_image, page_id)
 

ocrd_tesserocr/segment_region.py

@@ -26,6 +26,7 @@
     NoiseRegionType,
     to_xml)
 from ocrd_models.ocrd_page_generateds import TableRegionType
+from ocrd_models import OcrdExif
 from ocrd import Processor
 
 from .config import TESSDATA_PREFIX, OCRD_TOOL
@@ -126,6 +127,12 @@ def process(self):
                     else:
                         LOG.warning('keeping existing ReadingOrder')
                 page_image = self.workspace.resolve_image_as_pil(page.imageFilename)
+                page_image_info = OcrdExif(page_image)
+                if page_image_info.xResolution != 1:
+                    dpi = page_image_info.xResolution
+                    if page_image_info.resolutionUnit == 'cm':
+                        dpi = round(dpi * 2.54)
+                    tessapi.SetVariable('user_defined_dpi', str(dpi))
                 page_image, page_xywh = image_from_page(
                     self.workspace, page, page_image, page_id)
                 LOG.info("Detecting regions in page '%s'", page_id)

ocrd_tesserocr/segment_word.py

@@ -17,6 +17,7 @@
     WordType,
     to_xml
 )
+from ocrd_models import OcrdExif
 
 from ocrd_tesserocr.config import TESSDATA_PREFIX, OCRD_TOOL
 from .common import (
@@ -71,6 +72,12 @@ def process(self):
                                                                for name in self.parameter.keys()])]))
                 page = pcgts.get_Page()
                 page_image = self.workspace.resolve_image_as_pil(page.imageFilename)
+                page_image_info = OcrdExif(page_image)
+                if page_image_info.xResolution != 1:
+                    dpi = page_image_info.xResolution
+                    if page_image_info.resolutionUnit == 'cm':
+                        dpi = round(dpi * 2.54)
+                    tessapi.SetVariable('user_defined_dpi', str(dpi))
                 page_image, page_xywh = image_from_page(
                     self.workspace, page, page_image, page_id)