Experimental add jbig2

It appears that fitz forces conversion of jbig2 to ccitt no matter what,
so pikepdf will be needed to patch jbig2 images.

src/ocrmypdf/exec/jbig2enc.py

@@ -0,0 +1,39 @@
+# © 2018 James R. Barlow: github.com/jbarlow83
+#
+# This file is part of OCRmyPDF.
+#
+# OCRmyPDF is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# OCRmyPDF is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with OCRmyPDF.  If not, see <http://www.gnu.org/licenses/>.
+
+from subprocess import CalledProcessError, run
+from tempfile import TemporaryFile
+import sys
+import os
+import shutil
+
+from ..exceptions import ExitCode
+
+
+def convert(input_file, output_file):
+    args_jbig2 = [
+        'jbig2',
+        '-s',
+        input_file
+    ]
+
+    with TemporaryFile(mode='w+b') as tmp:
+        proc = run(args_jbig2, stdout=tmp)
+        
+        if proc.returncode == 0:
+            with open(output_file, 'wb') as output:
+                shutil.copyfileobj(tmp, output)

src/ocrmypdf/pipeline.py

@@ -803,6 +803,60 @@ def combine_layers(
         pdf_output.write(out)
 
 
+def jbig2_10to1(
+        infiles,
+        output_file,
+        log,
+        context):
+    assert fitz
+    options = context.get_options()
+    infiles = list(flatten_groups(infiles))
+
+    doc = fitz.open()
+    for infile in infiles:
+        subdoc = fitz.open(infile)
+        doc.insertPDF(subdoc)
+    
+    root = Path(output_file).parent / Path(output_file).stem
+    root.mkdir()
+    
+    changed_xrefs = []
+    for pageno in range(len(infiles)):
+        images = doc.getPageImageList(pageno)
+        for image in images:
+            xref, smask, w, h, bpc, cs, alt_cs, name, filt = image
+            if bpc != 1:
+                continue
+            pix = fitz.Pixmap(doc, xref)
+            pix.writePNG(str(root / '{:08d}.png'.format(xref)), savealpha=False)
+            changed_xrefs.append(xref)
+    doc.save(output_file + '_tmp.pdf')
+    doc = fitz.open(output_file + '_tmp.pdf')
+
+    from subprocess import run, PIPE
+    args = ['jbig2', '-s', '-p', '-v', '-S']
+    args.extend(root.glob('*.png'))
+    proc = run(args, cwd=str(root), stdout=PIPE, stderr=PIPE)
+    proc.check_returncode()
+    log.debug(proc.stderr)
+
+    jbig2_globals_xref = doc._getNewXref()
+    log.info(jbig2_globals_xref)
+    doc._updateStream(xref, (root / 'output.sym').read_bytes())
+
+    for n, xref in enumerate(changed_xrefs):
+        jbig2_im_file = root / 'output.{:04d}'.format(n)
+        obj_str = doc._getObjectString(xref)
+        log.info(xref)
+        log.info(obj_str)
+        obj_str = obj_str.replace('/Filter/FlateDecode', '/Filter/JBIG2Decode')
+        obj_str = obj_str.replace(
+            '>>', '/DecodeParms << /JBIG2Globals {} 0 R >> >>'.format(jbig2_globals_xref))
+        doc._updateObject(xref, obj_str)
+        doc._updateStream(xref, jbig2_im_file.read_bytes())
+    doc.save(output_file, garbage=0, deflate=0, clean=0, expand=0)
+
+
 def ocr_tesseract_and_render_pdf(
         infiles,
         outfiles,
@@ -1280,6 +1334,16 @@ def build_pipeline(options, work_folder, log, context):
     task_combine_layers.active_if(options.pdf_renderer == 'hocr' or 
                                   options.pdf_renderer == 'sandwich')
 
+    task_jbig2_10to1 = main_pipeline.collate(
+        task_func=jbig2_10to1,
+        input=[combine_layers],
+        filter=regex(r".*/(\d{5})(\d)(?:\.rendered\.pdf)"),
+        output=os.path.join(work_folder, r'\1x.opt.pdf'),
+        extras=[log, context])
+    task_jbig2_10to1.graphviz(fillcolor='"#00cc66"')
+    task_jbig2_10to1.active_if(fitz and (options.pdf_renderer == 'hocr' or 
+                                         options.pdf_renderer == 'sandwich'))
+
     # Tesseract OCR+PDF
     task_ocr_tesseract_and_render_pdf = main_pipeline.collate(
         task_func=ocr_tesseract_and_render_pdf,