@@ -437,7 +437,7 @@ def _qimage_to_pil(self, qimage: QImage) -> PIL.Image.Image:
437437return img .convert ("RGB" )
438438
439439def _ocr_image (self ,qimage :QImage ,langs :Optional [str ]= None ,timeout :int = 30 )-> bytes :
440- if shutil .which ("tesseract" )is None :
440+ if shutil .which (self . editor . args . tesseract_command )is None :
441441# tesseract is not installed
442442return None
443443langs = langs or self .editor .ocr_langs
@@ -448,14 +448,22 @@ def _ocr_image(self, qimage: QImage, langs: Optional[str] = None, timeout: int =
448448# TODO? use https://github.com/sirfz/tesserocr
449449tiff_bytes = pil_to_tiff_bytes (pil_img )
450450args = [
451- "tesseract" ,
451+ self . editor . args . tesseract_command ,
452452"-" ,# input: stdin
453453"-" ,# output: stdout
454454"-l" ,langs ,
455- # "-c", "tessedit_create_hocr=1", # config
456- "quiet" ,# config: hide "Estimating resolution as N" messages
457- "hocr" ,# extension
455+ "-c" ,"tessedit_create_hocr=1" ,
456+ # TODO get dpi value from hocr file
457+ # <div class='ocr_page' id='page_1' title='...; scan_res 300 300'>
458+ # "--dpi", "300",
459+ "--loglevel" ,"WARN" ,# ALL, TRACE, DEBUG, INFO, WARN, ERROR, FATAL, OFF
458460 ]
461+ if self .editor .args .tessdata_dir :
462+ args += [
463+ "--oem" ,"1" ,
464+ "--psm" ,"6" ,
465+ "--tessdata-dir" ,self .editor .args .tessdata_dir ,
466+ ]
459467hocr_bytes = subprocess .check_output (args ,input = tiff_bytes ,timeout = timeout )
460468return hocr_bytes
461469
@@ -1100,6 +1108,15 @@ def main():
11001108default = None ,
11011109help = "Overlay color (color name or #RRGGBB)" ,
11021110 )
1111+ parser .add_argument (
1112+ "--tesseract-command" ,
1113+ default = "tesseract" ,
1114+ )
1115+ parser .add_argument (
1116+ "--tessdata-dir" ,
1117+ default = None ,
1118+ help = "usually a git clone of https://github.com/tesseract-ocr/tessdata_best" ,
1119+ )
11031120args = parser .parse_args ()
11041121
11051122# handle Ctrl+C from terminal