# SPDX-FileCopyrightText: 2022 James R. Barlow # SPDX-License-Identifier: MPL-2.0 """Post-processing image optimization of OCR PDFs.""" from __future__ import annotations import logging import sys import tempfile import threading from collections import defaultdict from collections.abc import Callable, Iterator, MutableSet, Sequence from os import fspath from pathlib import Path from typing import Any, NamedTuple, NewType from zlib import compress import img2pdf from pikepdf import ( Dictionary, Name, Object, ObjectStreamMode, Pdf, PdfError, PdfImage, Stream, UnsupportedImageTypeError, ) from pikepdf.models.image import HifiPrintImageNotTranscodableError from PIL import Image from ocrmypdf._concurrent import Executor, SerialExecutor from ocrmypdf._exec import jbig2enc, pngquant from ocrmypdf._jobcontext import PdfContext from ocrmypdf._progressbar import ProgressBar from ocrmypdf.exceptions import OutputFileAccessError from ocrmypdf.helpers import IMG2PDF_KWARGS, safe_symlink log = logging.getLogger(__name__) DEFAULT_JPEG_QUALITY = 75 DEFAULT_PNG_QUALITY = 70 Xref = NewType('Xref', int) class XrefExt(NamedTuple): """A PDF xref and image extension pair.""" xref: Xref ext: str def img_name(root: Path, xref: Xref, ext: str) -> Path: """Return the name of an image file for a given xref and extension.""" return root / f'{xref:08d}{ext}' def png_name(root: Path, xref: Xref) -> Path: """Return the name of a PNG file for a given xref.""" return img_name(root, xref, '.png') def jpg_name(root: Path, xref: Xref) -> Path: """Return the name of a JPEG file for a given xref.""" return img_name(root, xref, '.jpg') def extract_image_filter( image: Stream, xref: Xref ) -> tuple[PdfImage, tuple[Name, Object]] | None: """Determine if an image is extractable.""" if image.Subtype != Name.Image: return None if not isinstance(image.Length, int) or image.Length < 100: log.debug(f"xref {xref}: skipping image with small stream size") return None if ( not isinstance(image.Width, int) or not isinstance(image.Height, int) or image.Width < 8 or image.Height < 8 ): # Issue 732 log.debug(f"xref {xref}: skipping image with unusually small dimensions") return None pim = PdfImage(image) if len(pim.filter_decodeparms) > 1: first_filtdp = pim.filter_decodeparms[0] second_filtdp = pim.filter_decodeparms[1] if ( len(pim.filter_decodeparms) == 2 and first_filtdp[0] == Name.FlateDecode and first_filtdp[1] is not None and first_filtdp[1].get(Name.Predictor, 1) == 1 and second_filtdp[0] == Name.DCTDecode and not second_filtdp[1] ): log.debug( f"xref {xref}: found image compressed as /FlateDecode /DCTDecode, " "marked for JPEG optimization" ) filtdp = pim.filter_decodeparms[1] else: log.debug(f"xref {xref}: skipping image with multiple compression filters") return None else: filtdp = pim.filter_decodeparms[0] if pim.bits_per_component > 8: log.debug(f"xref {xref}: skipping wide gamut image") return None # Don't mess with wide gamut images if filtdp[0] == Name.JPXDecode: log.debug(f"xref {xref}: skipping JPEG2000 image") return None # Don't do JPEG2000 if filtdp[0] == Name.CCITTFaxDecode and filtdp[1].get('/K', 0) >= 0: log.debug(f"xref {xref}: skipping CCITT Group 3 image") return None # pikepdf doesn't support Group 3 yet if Name.Decode in image: log.debug(f"xref {xref}: skipping image with Decode table") return None # Don't mess with custom Decode tables return pim, filtdp def extract_image_jbig2( *, pdf: Pdf, root: Path, image: Stream, xref: Xref, options ) -> XrefExt | None: """Extract an image, saving it as a JBIG2 file.""" del options # unused arg result = extract_image_filter(image, xref) if result is None: return None pim, filtdp = result if ( pim.bits_per_component == 1 and filtdp[0] != Name.JBIG2Decode and jbig2enc.available() ): # Save any colorspace associated with the image, so that we # will export a pure 1-bit PNG with no palette or ICC profile. # Showing the palette or ICC to jbig2enc will cause it to perform # colorspace transform to 1bpp, which will conflict the palette or # ICC if it exists. colorspace = pim.obj.get(Name.ColorSpace, None) if colorspace is not None or pim.image_mask: try: # Set to DeviceGray temporarily; we already in 1 bpc. pim.obj.ColorSpace = Name.DeviceGray imgname = root / f'{xref:08d}' with imgname.open('wb') as f: ext = pim.extract_to(stream=f) # Rename the file so it has .prejbig2.ext extension # Making it unique avoids problems with Windows if the # same image is extracted multiple times imgname.rename(imgname.with_suffix(".prejbig2" + ext)) except NotImplementedError as e: if '/Decode' in str(e): log.debug( f"xref {xref}: skipping image with unsupported Decode table" ) return None raise except UnsupportedImageTypeError: return None finally: # Restore image colorspace after temporarily setting it to DeviceGray if colorspace is not None: pim.obj.ColorSpace = colorspace else: del pim.obj.ColorSpace return XrefExt(xref, ".prejbig2" + ext) return None def extract_image_generic( *, pdf: Pdf, root: Path, image: Stream, xref: Xref, options ) -> XrefExt | None: """Generic image extraction.""" result = extract_image_filter(image, xref) if result is None: return None pim, filtdp = result # Don't try to PNG-optimize 1bpp images, since JBIG2 does it better. if pim.bits_per_component == 1: return None if filtdp[0] == Name.DCTDecode and options.optimize >= 2: # This is a simple heuristic derived from some training data, that has # about a 70% chance of guessing whether the JPEG is high quality, # and possibly recompressible, or not. The number itself doesn't mean # anything. # bytes_per_pixel = int(raw_jpeg.Length) / (w * h) # jpeg_quality_estimate = 117.0 * (bytes_per_pixel ** 0.213) # if jpeg_quality_estimate < 65: # return None try: imgname = root / f'{xref:08d}' with imgname.open('wb') as f: ext = pim.extract_to(stream=f) imgname.rename(imgname.with_suffix(ext)) except (UnsupportedImageTypeError, HifiPrintImageNotTranscodableError): return None return XrefExt(xref, ext) elif ( pim.indexed and pim.colorspace in pim.SIMPLE_COLORSPACES and options.optimize >= 3 ): # Try to improve on indexed images - these are far from low hanging # fruit in most cases pim.as_pil_image().save(png_name(root, xref)) return XrefExt(xref, '.png') elif not pim.indexed and pim.colorspace in pim.SIMPLE_COLORSPACES: # An optimization opportunity here, not currently taken, is directly # generating a PNG from compressed data try: pim.as_pil_image().save(png_name(root, xref)) except NotImplementedError: log.warning("PDF contains an atypical image that cannot be optimized.") return None return XrefExt(xref, '.png') elif ( not pim.indexed and pim.colorspace == Name.ICCBased and pim.bits_per_component == 1 and not options.jbig2_lossy ): # We can losslessly optimize 1-bit images to CCITT or JBIG2 without # paying any attention to the ICC profile, provided we're not doing # lossy JBIG2 pim.as_pil_image().save(png_name(root, xref)) return XrefExt(xref, '.png') return None def _find_image_xrefs_container( pdf: Pdf, container: Object, pageno: int, include_xrefs: MutableSet[Xref], exclude_xrefs: MutableSet[Xref], pageno_for_xref: dict[Xref, int], depth: int = 0, ): """Find all image XRefs or Form XObject and add to the include/exclude sets.""" if depth > 10: log.warning("Recursion depth exceeded in _find_image_xrefs_page") return try: xobjs = container.Resources.XObject except AttributeError: return for _imname, image in dict(xobjs).items(): if image.objgen[1] != 0: continue # Ignore images in an incremental PDF xref = Xref(image.objgen[0]) if xref in include_xrefs or xref in exclude_xrefs: continue # Already processed if Name.Subtype in image and image.Subtype == Name.Form: # Recurse into Form XObjects log.debug(f"Recursing into Form XObject {_imname} in page {pageno}") _find_image_xrefs_container( pdf, image, pageno, include_xrefs, exclude_xrefs, pageno_for_xref, depth + 1, ) continue if Name.SMask in image: # Ignore soft masks smask_xref = Xref(image.SMask.objgen[0]) exclude_xrefs.add(smask_xref) log.debug(f"xref {smask_xref}: skipping image because it is an SMask") include_xrefs.add(xref) log.debug(f"xref {xref}: treating as an optimization candidate") if xref not in pageno_for_xref: pageno_for_xref[xref] = pageno def _find_image_xrefs(pdf: Pdf): include_xrefs: MutableSet[Xref] = set() exclude_xrefs: MutableSet[Xref] = set() pageno_for_xref: dict[Xref, int] = {} for pageno, page in enumerate(pdf.pages): _find_image_xrefs_container( pdf, page.obj, pageno, include_xrefs, exclude_xrefs, pageno_for_xref ) working_xrefs = include_xrefs - exclude_xrefs return working_xrefs, pageno_for_xref def extract_images( pdf: Pdf, root: Path, options, extract_fn: Callable[..., XrefExt | None], ) -> Iterator[tuple[int, XrefExt]]: """Extract image using extract_fn. Enumerate images on each page, lookup their xref/ID number in the PDF. Exclude images that are soft masks (i.e. alpha transparency related). Record the page number on which an image is first used, since images may be used on multiple pages (or multiple times on the same page). Current we do not check Form XObjects or other objects that may contain images, and we don't evaluate alternate images or thumbnails. extract_fn must decide if wants to extract the image in this context. If it does a tuple should be returned: (xref, ext) where .ext is the file extension. extract_fn must also extract the file it finds interesting. """ errors = 0 working_xrefs, pageno_for_xref = _find_image_xrefs(pdf) for xref in working_xrefs: image = pdf.get_object((xref, 0)) try: result = extract_fn( pdf=pdf, root=root, image=image, xref=xref, options=options ) except Exception: # pylint: disable=broad-except log.exception( f"xref {xref}: While extracting this image, an error occurred" ) errors += 1 else: if result: _, ext = result yield pageno_for_xref[xref], XrefExt(xref, ext) def extract_images_generic( pdf: Pdf, root: Path, options ) -> tuple[list[Xref], list[Xref]]: """Extract any >=2bpp image we think we can improve.""" jpegs = [] pngs = [] for _, xref_ext in extract_images(pdf, root, options, extract_image_generic): log.debug('%s', xref_ext) if xref_ext.ext == '.png': pngs.append(xref_ext.xref) elif xref_ext.ext == '.jpg': jpegs.append(xref_ext.xref) log.debug(f"Optimizable images: JPEGs: {len(jpegs)} PNGs: {len(pngs)}") return jpegs, pngs def extract_images_jbig2(pdf: Pdf, root: Path, options) -> dict[int, list[XrefExt]]: """Extract any bitonal image that we think we can improve as JBIG2.""" jbig2_groups = defaultdict(list) for pageno, xref_ext in extract_images(pdf, root, options, extract_image_jbig2): group = pageno // options.jbig2_page_group_size jbig2_groups[group].append(xref_ext) log.debug(f"Optimizable images: JBIG2 groups: {len(jbig2_groups)}") return jbig2_groups def _produce_jbig2_images( jbig2_groups: dict[int, list[XrefExt]], root: Path, options, executor: Executor ) -> None: """Produce JBIG2 images from their groups.""" def jbig2_group_args(root: Path, groups: dict[int, list[XrefExt]]): for group, xref_exts in groups.items(): prefix = f'group{group:08d}' yield ( fspath(root), # =cwd (img_name(root, xref, ext) for xref, ext in xref_exts), # =infiles prefix, # =out_prefix options.jbig2_threshold, ) def jbig2_single_args(root: Path, groups: dict[int, list[XrefExt]]): for group, xref_exts in groups.items(): prefix = f'group{group:08d}' # Second loop is to ensure multiple images per page are unpacked for n, xref_ext in enumerate(xref_exts): xref, ext = xref_ext yield ( fspath(root), img_name(root, xref, ext), root / f'{prefix}.{n:04d}', options.jbig2_threshold, ) if options.jbig2_page_group_size > 1: jbig2_args = jbig2_group_args jbig2_convert = jbig2enc.convert_group else: jbig2_args = jbig2_single_args jbig2_convert = jbig2enc.convert_single executor( use_threads=True, max_workers=options.jobs, progress_kwargs=dict( total=len(jbig2_groups), desc="JBIG2", unit='item', disable=not options.progress_bar, ), task=jbig2_convert, task_arguments=jbig2_args(root, jbig2_groups), ) def convert_to_jbig2( pdf: Pdf, jbig2_groups: dict[int, list[XrefExt]], root: Path, options, executor: Executor, ) -> None: """Convert images to JBIG2 and insert into PDF. When the JBIG2 page group size is > 1 we do several JBIG2 images at once and build a symbol dictionary that will span several pages. Each JBIG2 image must reference to its symbol dictionary. If too many pages shared the same dictionary JBIG2 encoding becomes more expensive and less efficient. The default value of 10 was determined through testing. Currently this must be lossy encoding since jbig2enc does not support refinement coding. When the JBIG2 symbolic coder is not used, each JBIG2 stands on its own and needs no dictionary. Currently this must be lossless JBIG2. """ jbig2_globals_dict: Dictionary | None _produce_jbig2_images(jbig2_groups, root, options, executor) for group, xref_exts in jbig2_groups.items(): prefix = f'group{group:08d}' jbig2_symfile = root / (prefix + '.sym') if jbig2_symfile.exists(): jbig2_globals_data = jbig2_symfile.read_bytes() jbig2_globals = Stream(pdf, jbig2_globals_data) jbig2_globals_dict = Dictionary(JBIG2Globals=jbig2_globals) elif options.jbig2_page_group_size == 1: jbig2_globals_dict = None else: raise FileNotFoundError(jbig2_symfile) for n, xref_ext in enumerate(xref_exts): xref, _ = xref_ext jbig2_im_file = root / (prefix + f'.{n:04d}') jbig2_im_data = jbig2_im_file.read_bytes() im_obj = pdf.get_object(xref, 0) im_obj.write( jbig2_im_data, filter=Name.JBIG2Decode, decode_parms=jbig2_globals_dict ) def _optimize_jpeg( xref: Xref, in_jpg: Path, opt_jpg: Path, jpeg_quality: int ) -> tuple[Xref, Path | None]: with Image.open(in_jpg) as im: im.save(opt_jpg, optimize=True, quality=jpeg_quality) if opt_jpg.stat().st_size > in_jpg.stat().st_size: log.debug(f"xref {xref}, jpeg, made larger - skip") opt_jpg.unlink() return xref, None return xref, opt_jpg def transcode_jpegs( pdf: Pdf, jpegs: Sequence[Xref], root: Path, options, executor: Executor ) -> None: """Optimize JPEGs according to optimization settings.""" def jpeg_args() -> Iterator[tuple[Xref, Path, Path, int]]: for xref in jpegs: in_jpg = jpg_name(root, xref) opt_jpg = in_jpg.with_suffix('.opt.jpg') yield xref, in_jpg, opt_jpg, options.jpeg_quality def finish_jpeg(result: tuple[Xref, Path | None], pbar: ProgressBar): xref, opt_jpg = result if opt_jpg: compdata = opt_jpg.read_bytes() # JPEG can inserted into PDF as is im_obj = pdf.get_object(xref, 0) im_obj.write(compdata, filter=Name.DCTDecode) pbar.update() executor( use_threads=True, # Processes are significantly slower at this task max_workers=options.jobs, progress_kwargs=dict( desc="Recompressing JPEGs", total=len(jpegs), unit='image', disable=not options.progress_bar, ), task=_optimize_jpeg, task_arguments=jpeg_args(), task_finished=finish_jpeg, ) def _find_deflatable_jpeg( *, pdf: Pdf, root: Path, image: Stream, xref: Xref, options ) -> XrefExt | None: result = extract_image_filter(image, xref) if result is None: return None _pim, filtdp = result if filtdp[0] == Name.DCTDecode and not filtdp[1] and options.optimize >= 1: return XrefExt(xref, '.memory') return None def _deflate_jpeg( pdf: Pdf, lock: threading.Lock, xref: Xref, complevel: int ) -> tuple[Xref, bytes]: with lock: xobj = pdf.get_object(xref, 0) try: data = xobj.read_raw_bytes() except PdfError: return xref, b'' compdata = compress(data, complevel) if len(compdata) >= len(data): return xref, b'' return xref, compdata def deflate_jpegs(pdf: Pdf, root: Path, options, executor: Executor) -> None: """Apply FlateDecode to JPEGs. This is a lossless compression method that is supported by all PDF viewers, and generally results in a smaller file size compared to straight DCTDecode images. """ jpegs = [] for _pageno, xref_ext in extract_images(pdf, root, options, _find_deflatable_jpeg): xref = xref_ext.xref log.debug(f'xref {xref}: marking this JPEG as deflatable') jpegs.append(xref) complevel = 9 if options.optimize == 3 else 6 # Our calls to xobj.write() in finish() need coordination lock = threading.Lock() def deflate_args() -> Iterator: for xref in jpegs: yield pdf, lock, xref, complevel def finish(result: tuple[Xref, bytes], pbar: ProgressBar): xref, compdata = result if len(compdata) > 0: with lock: xobj = pdf.get_object(xref, 0) xobj.write(compdata, filter=[Name.FlateDecode, Name.DCTDecode]) pbar.update() executor( use_threads=True, # We're sharing the pdf directly, must use threads max_workers=options.jobs, progress_kwargs=dict( desc="Deflating JPEGs", total=len(jpegs), unit='image', disable=not options.progress_bar, ), task=_deflate_jpeg, task_arguments=deflate_args(), task_finished=finish, ) def _transcode_png(pdf: Pdf, filename: Path, xref: Xref) -> bool: output = filename.with_suffix('.png.pdf') with output.open('wb') as f: img2pdf.convert(fspath(filename), outputstream=f, **IMG2PDF_KWARGS) with Pdf.open(output) as pdf_image: foreign_image = next(iter(pdf_image.pages[0].images.values())) local_image = pdf.copy_foreign(foreign_image) im_obj = pdf.get_object(xref, 0) im_obj.write( local_image.read_raw_bytes(), filter=local_image.Filter, decode_parms=local_image.DecodeParms, ) # Don't copy keys from the new image... del_keys = set(im_obj.keys()) - set(local_image.keys()) # ...except for the keep_fields, which are essential to displaying # the image correctly and preserving its metadata. (/Decode arrays # and /SMaskInData are implicitly discarded prior to this point.) keep_fields = { '/ID', '/Intent', '/Interpolate', '/Mask', '/Metadata', '/OC', '/OPI', '/SMask', '/StructParent', } del_keys -= keep_fields for key in local_image.keys(): if key != Name.Length and str(key) not in keep_fields: im_obj[key] = local_image[key] for key in del_keys: del im_obj[key] return True def transcode_pngs( pdf: Pdf, images: Sequence[Xref], image_name_fn: Callable[[Path, Xref], Path], root: Path, options, executor: Executor, ) -> None: """Apply lossy transcoding to PNGs.""" modified: MutableSet[Xref] = set() if options.optimize >= 2: png_quality = ( max(10, options.png_quality - 10), min(100, options.png_quality + 10), ) def pngquant_args(): for xref in images: log.debug(image_name_fn(root, xref)) yield ( image_name_fn(root, xref), png_name(root, xref), png_quality[0], png_quality[1], ) modified.add(xref) executor( use_threads=True, max_workers=options.jobs, progress_kwargs=dict( desc="PNGs", total=len(images), unit='image', disable=not options.progress_bar, ), task=pngquant.quantize, task_arguments=pngquant_args(), ) for xref in modified: filename = png_name(root, xref) _transcode_png(pdf, filename, xref) DEFAULT_EXECUTOR = SerialExecutor() def optimize( input_file: Path, output_file: Path, context: PdfContext, save_settings: dict[str, Any], executor: Executor = DEFAULT_EXECUTOR, ) -> Path: """Optimize images in a PDF file.""" options = context.options if options.optimize == 0: safe_symlink(input_file, output_file) return output_file if options.jpeg_quality == 0: options.jpeg_quality = DEFAULT_JPEG_QUALITY if options.optimize < 3 else 40 if options.png_quality == 0: options.png_quality = DEFAULT_PNG_QUALITY if options.optimize < 3 else 30 if options.jbig2_page_group_size == 0: options.jbig2_page_group_size = 10 if options.jbig2_lossy else 1 with Pdf.open(input_file) as pdf: root = output_file.parent / 'images' root.mkdir(exist_ok=True) jpegs, pngs = extract_images_generic(pdf, root, options) transcode_jpegs(pdf, jpegs, root, options, executor) deflate_jpegs(pdf, root, options, executor) # if options.optimize >= 2: # Try pngifying the jpegs # transcode_pngs(pdf, jpegs, jpg_name, root, options) transcode_pngs(pdf, pngs, png_name, root, options, executor) jbig2_groups = extract_images_jbig2(pdf, root, options) convert_to_jbig2(pdf, jbig2_groups, root, options, executor) target_file = output_file.with_suffix('.opt.pdf') pdf.remove_unreferenced_resources() pdf.save(target_file, **save_settings) input_size = input_file.stat().st_size output_size = target_file.stat().st_size if output_size == 0: raise OutputFileAccessError( f"Output file not created after optimizing. We probably ran " f"out of disk space in the temporary folder: {tempfile.gettempdir()}." ) savings = 1 - output_size / input_size if savings < 0: log.info( "Image optimization did not improve the file - " "optimizations will not be used" ) # We still need to save the file with Pdf.open(input_file) as pdf: pdf.remove_unreferenced_resources() pdf.save(output_file, **save_settings) else: safe_symlink(target_file, output_file) return output_file def main(infile, outfile, level, jobs=1): """Entry point for direct optimization of a file.""" from shutil import copy # pylint: disable=import-outside-toplevel from tempfile import TemporaryDirectory # pylint: disable=import-outside-toplevel class OptimizeOptions: """Emulate ocrmypdf's options.""" def __init__( self, input_file, jobs, optimize_, jpeg_quality, png_quality, jb2lossy ): self.input_file = input_file self.jobs = jobs self.optimize = optimize_ self.jpeg_quality = jpeg_quality self.png_quality = png_quality self.jbig2_page_group_size = 0 self.jbig2_lossy = jb2lossy self.jbig2_threshold = 0.85 self.quiet = True self.progress_bar = False infile = Path(infile) options = OptimizeOptions( input_file=infile, jobs=jobs, optimize_=int(level), jpeg_quality=0, # Use default png_quality=0, jb2lossy=False, ) with TemporaryDirectory() as tmpdir: context = PdfContext(options, tmpdir, infile, None, None) tmpout = Path(tmpdir) / 'out.pdf' optimize( infile, tmpout, context, dict( compress_streams=True, preserve_pdfa=True, object_stream_mode=ObjectStreamMode.generate, ), ) copy(fspath(tmpout), fspath(outfile)) if __name__ == '__main__': main(sys.argv[1], sys.argv[2], sys.argv[3])