forked from josch/img2pdf
Interlaced PNGs cannot be directly embedded but have to be re-encoded
This commit is contained in:
parent
7244d2c6ed
commit
d9a6c9db03
1 changed files with 70 additions and 58 deletions
128
src/img2pdf.py
128
src/img2pdf.py
|
@ -759,6 +759,8 @@ def read_images(rawdata, colorspace, first_frame_only=False):
|
||||||
|
|
||||||
# depending on the input format, determine whether to pass the raw
|
# depending on the input format, determine whether to pass the raw
|
||||||
# image or the zlib compressed color information
|
# image or the zlib compressed color information
|
||||||
|
|
||||||
|
# JPEG and JPEG2000 can be embedded into the PDF as-is
|
||||||
if imgformat == ImageFormat.JPEG or imgformat == ImageFormat.JPEG2000:
|
if imgformat == ImageFormat.JPEG or imgformat == ImageFormat.JPEG2000:
|
||||||
color, ndpi, imgwidthpx, imgheightpx = get_imgmetadata(
|
color, ndpi, imgwidthpx, imgheightpx = get_imgmetadata(
|
||||||
imgdata, imgformat, default_dpi, colorspace, rawdata)
|
imgdata, imgformat, default_dpi, colorspace, rawdata)
|
||||||
|
@ -770,71 +772,81 @@ def read_images(rawdata, colorspace, first_frame_only=False):
|
||||||
raise JpegColorspaceError("jpeg can't have an alpha channel")
|
raise JpegColorspaceError("jpeg can't have an alpha channel")
|
||||||
im.close()
|
im.close()
|
||||||
return [(color, ndpi, imgformat, rawdata, imgwidthpx, imgheightpx, [])]
|
return [(color, ndpi, imgformat, rawdata, imgwidthpx, imgheightpx, [])]
|
||||||
elif imgformat == ImageFormat.PNG:
|
|
||||||
|
# We can directly embed the IDAT chunk of PNG images if the PNG is not
|
||||||
|
# interlaced
|
||||||
|
#
|
||||||
|
# PIL does not provide the information whether a PNG was stored interlaced
|
||||||
|
# or not. Thus, we retrieve that info manually by looking at byte 13 in the
|
||||||
|
# IHDR chunk. We know where to find that in the file because the IHDR chunk
|
||||||
|
# must be the first chunk.
|
||||||
|
if imgformat == ImageFormat.PNG and rawdata[28] == 0:
|
||||||
color, ndpi, imgwidthpx, imgheightpx = get_imgmetadata(
|
color, ndpi, imgwidthpx, imgheightpx = get_imgmetadata(
|
||||||
imgdata, imgformat, default_dpi, colorspace, rawdata)
|
imgdata, imgformat, default_dpi, colorspace, rawdata)
|
||||||
pngidat, palette = parse_png(rawdata)
|
pngidat, palette = parse_png(rawdata)
|
||||||
return [(color, ndpi, imgformat, pngidat, imgwidthpx, imgheightpx, palette)]
|
return [(color, ndpi, imgformat, pngidat, imgwidthpx, imgheightpx, palette)]
|
||||||
else:
|
|
||||||
result = []
|
|
||||||
img_page_count = 0
|
|
||||||
# loop through all frames of the image (example: multipage TIFF)
|
|
||||||
while True:
|
|
||||||
try:
|
|
||||||
imgdata.seek(img_page_count)
|
|
||||||
except EOFError:
|
|
||||||
break
|
|
||||||
|
|
||||||
if first_frame_only and img_page_count > 0:
|
# Everything else has to be encoded
|
||||||
break
|
|
||||||
|
|
||||||
logging.debug("Converting frame: %d" % img_page_count)
|
result = []
|
||||||
|
img_page_count = 0
|
||||||
color, ndpi, imgwidthpx, imgheightpx = get_imgmetadata(
|
# loop through all frames of the image (example: multipage TIFF)
|
||||||
imgdata, imgformat, default_dpi, colorspace)
|
while True:
|
||||||
|
|
||||||
newimg = None
|
|
||||||
if color == Colorspace['1']:
|
|
||||||
try:
|
|
||||||
ccittdata = transcode_monochrome(imgdata)
|
|
||||||
imgformat = ImageFormat.CCITTGroup4
|
|
||||||
result.append((color, ndpi, imgformat, ccittdata,
|
|
||||||
imgwidthpx, imgheightpx))
|
|
||||||
img_page_count += 1
|
|
||||||
continue
|
|
||||||
except Exception as e:
|
|
||||||
logging.debug(e)
|
|
||||||
logging.debug("Converting colorspace 1 to L")
|
|
||||||
newimg = imgdata.convert('L')
|
|
||||||
color = Colorspace.L
|
|
||||||
elif color in [Colorspace.RGB, Colorspace.L, Colorspace.CMYK,
|
|
||||||
Colorspace["CMYK;I"], Colorspace.P]:
|
|
||||||
logging.debug("Colorspace is OK: %s", color)
|
|
||||||
newimg = imgdata
|
|
||||||
elif color in [Colorspace.RGBA, Colorspace.other]:
|
|
||||||
logging.debug("Converting colorspace %s to RGB", color)
|
|
||||||
newimg = imgdata.convert('RGB')
|
|
||||||
color = Colorspace.RGB
|
|
||||||
else:
|
|
||||||
raise ValueError("unknown colorspace: %s" % color.name)
|
|
||||||
# cheapo version to retrieve a PNG encoding of the payload is to
|
|
||||||
# just save it with PIL. In the future this could be replaced by
|
|
||||||
# dedicated function applying the Paeth PNG filter to the raw pixel
|
|
||||||
pngbuffer = BytesIO()
|
|
||||||
newimg.save(pngbuffer, format="png")
|
|
||||||
pngidat, palette = parse_png(pngbuffer.getvalue())
|
|
||||||
imgformat = ImageFormat.PNG
|
|
||||||
result.append((color, ndpi, imgformat, pngidat, imgwidthpx,
|
|
||||||
imgheightpx, palette))
|
|
||||||
img_page_count += 1
|
|
||||||
# the python-pil version 2.3.0-1ubuntu3 in Ubuntu does not have the
|
|
||||||
# close() method
|
|
||||||
try:
|
try:
|
||||||
imgdata.close()
|
imgdata.seek(img_page_count)
|
||||||
except AttributeError:
|
except EOFError:
|
||||||
pass
|
break
|
||||||
im.close()
|
|
||||||
return result
|
if first_frame_only and img_page_count > 0:
|
||||||
|
break
|
||||||
|
|
||||||
|
logging.debug("Converting frame: %d" % img_page_count)
|
||||||
|
|
||||||
|
color, ndpi, imgwidthpx, imgheightpx = get_imgmetadata(
|
||||||
|
imgdata, imgformat, default_dpi, colorspace)
|
||||||
|
|
||||||
|
newimg = None
|
||||||
|
if color == Colorspace['1']:
|
||||||
|
try:
|
||||||
|
ccittdata = transcode_monochrome(imgdata)
|
||||||
|
imgformat = ImageFormat.CCITTGroup4
|
||||||
|
result.append((color, ndpi, imgformat, ccittdata,
|
||||||
|
imgwidthpx, imgheightpx))
|
||||||
|
img_page_count += 1
|
||||||
|
continue
|
||||||
|
except Exception as e:
|
||||||
|
logging.debug(e)
|
||||||
|
logging.debug("Converting colorspace 1 to L")
|
||||||
|
newimg = imgdata.convert('L')
|
||||||
|
color = Colorspace.L
|
||||||
|
elif color in [Colorspace.RGB, Colorspace.L, Colorspace.CMYK,
|
||||||
|
Colorspace["CMYK;I"], Colorspace.P]:
|
||||||
|
logging.debug("Colorspace is OK: %s", color)
|
||||||
|
newimg = imgdata
|
||||||
|
elif color in [Colorspace.RGBA, Colorspace.other]:
|
||||||
|
logging.debug("Converting colorspace %s to RGB", color)
|
||||||
|
newimg = imgdata.convert('RGB')
|
||||||
|
color = Colorspace.RGB
|
||||||
|
else:
|
||||||
|
raise ValueError("unknown colorspace: %s" % color.name)
|
||||||
|
# cheapo version to retrieve a PNG encoding of the payload is to
|
||||||
|
# just save it with PIL. In the future this could be replaced by
|
||||||
|
# dedicated function applying the Paeth PNG filter to the raw pixel
|
||||||
|
pngbuffer = BytesIO()
|
||||||
|
newimg.save(pngbuffer, format="png")
|
||||||
|
pngidat, palette = parse_png(pngbuffer.getvalue())
|
||||||
|
imgformat = ImageFormat.PNG
|
||||||
|
result.append((color, ndpi, imgformat, pngidat, imgwidthpx,
|
||||||
|
imgheightpx, palette))
|
||||||
|
img_page_count += 1
|
||||||
|
# the python-pil version 2.3.0-1ubuntu3 in Ubuntu does not have the
|
||||||
|
# close() method
|
||||||
|
try:
|
||||||
|
imgdata.close()
|
||||||
|
except AttributeError:
|
||||||
|
pass
|
||||||
|
im.close()
|
||||||
|
return result
|
||||||
|
|
||||||
|
|
||||||
# converts a length in pixels to a length in PDF units (1/72 of an inch)
|
# converts a length in pixels to a length in PDF units (1/72 of an inch)
|
||||||
|
|
Loading…
Reference in a new issue