A tag tree is a data structure used for deserializing JPEG2000
packet headers.
We don't use them for anything yet, except from tests.
The implementation feels a bit awkward to me, but we can always polish
it later.
The spec thankfully includes two concrete examples. The code is
correct enough to pass those -- I added them as test.
The implementation is very similar to #23831.
I created the test exactly like in #23713, except that I replaced the
last four lines in the ini file with:
```
-txt -Param -rATX1 10
-txt -Param -rATY1 -1
-txt -Param -rATX2 4
-txt -Param -rATY2 15
```
This needed the same `jbig2` changes as for the non-transposed ones,
and the changes to it mentioned on #23780.
I used the same .ini files as for the non-transposed ones, except
that I added `-txt -Param -Transposed 1` as last line to each of them.
All three new files display fine in Chrome.
They all look busted in Firefox.
I think this is likey a bug in pdf.js that I'll report upstream.
(Reportedly they look fine in Acrobat on Android.)
This already worked fine. Now it's tested.
I did have to teach `jbig2` to correctly generate test files for this.
See the PR adding these tests for local changes.
I used the script from #23659 to create these images, but I replaced
these lines:
```
-txt -Param -numInst 4
-ID 2 108 50 -ID 3 265 60 -ID 1 100 135 -ID 0 70 232
-txt -Param -RefCorner 2
```
For `bottomleft`, I replaced them with:
```
-txt -Param -numInst 4
-ID 2 137 50 -ID 3 294 60 -ID 1 199 135 -ID 0 319 232
-txt -Param -RefCorner 0
```
For `bottomright`, I replaced them with:
```
-txt -Param -numInst 4
-ID 2 108 50 -ID 3 265 60 -ID 1 100 135 -ID 0 70 232
-txt -Param -RefCorner 2
```
For `topright`, I replaced them with:
```
-txt -Param -numInst 4
-ID 2 108 79 -ID 3 265 89 -ID 1 100 234 -ID 0 70 351
-txt -Param -RefCorner 3
```
All three new files display fine in Chrome.
The bottomleft one displays fine in Firefox, while the other two
look compressed in X. I think this is a bug in pdf.js that I'll
report upstream.
(Reportedly they look fine in Acrobat on Android.)
See the PR adding this test for local changes to `jbig2`.
I used the shell script mentioned in #23659, except I added the line
`-txt -Param -Transposed 1` at the very end of the .ini file.
As with all the symbol test cases, after running
Meta/jbig2_to_pdf.py -o foo.pdf foo.jb2 399 400
the file opens up ok in Chrome and Firefox (but not Safari), so
maybe it's not completely broken.
The T.800 spec says there should only be one 'colr' box, but the
extended jpx file format spec in T.801 annex M allows having multiple.
Method 2 is a basic ICC profile, while method 3 (jpx-only) allows full
ICC profiles. Support that.
For the test, I opened buggie.png in Photoshop, converted it to
grayscale, and saved it as a JPEG2000, with "JP2 Compatible" checked
and "Include Transparency" unchecked. I also unchecked "Include
Metadata", and "Lossless". I left "Fast Mode" checked and the quality
at the default 50.
This adds a test for the code added in #23710.
I created this file using `jbig2` (see below for details), but as
usual it required a bunch of changes to it to make it actually produce
spec-compliant output. See the PR adding this image for my local diff.
I created the test image file by running this shell script with
`jbig2` tweaked as described above:
#!/bin/bash
set -eu
S=Tests/LibGfx/test-inputs/bmp/bitmap.bmp
# See make-symbol-jbig.sh (the script in #23659) for the general
# setup and some comments. See also make-symbol-textrefine.sh (in
# #23713).
#
# `-Ref` takes 5 arguments:
# 1. The symbol ID of this symbol (like after a `-Simple`)
# 2. A bmp file that the base symbol gets refined to
# 3. The ID of the base symbol
# 4. dx, dy
cat << EOF > jbig2-symbol-symbolrefine.ini
-sym -Seg 1
-sym -file -numClass -HeightClass 3 -WidthClass 1
-sym -file -numSymbol 3
-sym -file -Height 250
-sym -file -Width 120 -Simple 0 mouth-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -file -Height 100
-sym -file -Width 100 -Simple 1 nose-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -file -Height 30
-sym -file -Width 30 -Simple 2 top_eye-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -Param -Huff_DH 0
-sym -Param -Huff_DW 0
-sym -Seg 2
-sym -file -numClass -HeightClass 1 -WidthClass 1
-sym -file -numSymbol 1
-sym -file -Height 30
-sym -file -Width 30 -Ref 3 bottom_eye-1bpp.bmp 2 0 0
-sym -file -EndOfHeightClass
-sym -Param -Huff_DH 0
-sym -Param -Huff_DW 0
-sym -Param -RefTemplate 1
-txt -Seg 3
-txt -Param -numInst 4
-ID 2 108 50 -ID 3 265 60 -ID 1 100 135 -ID 0 70 232
-txt -Param -RefCorner 1
-txt -Param -Xlocation 0
-txt -Param -Ylocation 0
-txt -Param -W 399
-txt -Param -H 400
EOF
J=$HOME/Downloads/T-REC-T.88-201808-I\!\!SOFT-ZST-E/Software
J=$J/JBIG2_SampleSoftware-A20180829/source/jbig2
$J -i "${S%.bmp}" -f bmp -o symbol-symbolrefine -F jb2 \
-ini jbig2-symbol-symbolrefine.ini
We can't decode any actual image data yet, but it shows that we can
read the basics of the container format. (...as long as there's an
Annex I container around the data, not just an Annex A codestream.
All files I've found so far have the container.)
I drew the thes input in Acorn.app and used "Save as..." to save it as
JPEG2000. It's an RGBA image.
This adds a test for the code added in #23696.
I created this file using `jbig2` (see below for details), but as
usual it required a bunch of changes to it to make it actually produce
spec-compliant output. See the PR adding this image for my local diff.
I created the test image file by running this shell script with
`jbig2` tweaked as described above:
#!/bin/bash
set -eu
S=Tests/LibGfx/test-inputs/bmp/bitmap.bmp
# See make-symbol-jbig.sh (the script in #23659) for the general
# setup and some comments. Note that the symbol section here only
# has 3 symbols, instead of 4 over there.
#
# `-RefID` takes 6 arguments:
# 1. The symbol ID of the base symbol (like after an `-ID`)
# 2. A bmp file that the base symbol gets refined to
# 3. y, x (like after an `-ID`)
# 4. dx, dy (note swapped order to previous item)
#
# We also explicitly set refinement adaptive pixels, because the
# default adaptive refinement pixels aren't the nominal pixels from
# the spec.
cat << EOF > jbig2-symbol-textrefine.ini
-sym -Seg 1
-sym -file -numClass -HeightClass 3 -WidthClass 1
-sym -file -numSymbol 3
-sym -file -Height 250
-sym -file -Width 120 -Simple 0 mouth-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -file -Height 100
-sym -file -Width 100 -Simple 1 nose-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -file -Height 30
-sym -file -Width 30 -Simple 2 top_eye-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -Param -Huff_DH 0
-sym -Param -Huff_DW 0
-txt -Seg 2
-txt -Param -numInst 4
-ID 2 108 50 -RefID 2 bottom_eye-1bpp.bmp 265 60 0 0
-ID 1 100 135 -ID 0 70 232
-txt -Param -RefCorner 1
-txt -Param -Xlocation 0
-txt -Param -Ylocation 0
-txt -Param -W 399
-txt -Param -H 400
-txt -Param -rATX1 -1
-txt -Param -rATY1 -1
-txt -Param -rATX2 -1
-txt -Param -rATY2 -1
EOF
J=$HOME/Downloads/T-REC-T.88-201808-I\!\!SOFT-ZST-E/Software
J=$J/JBIG2_SampleSoftware-A20180829/source/jbig2
$J -i "${S%.bmp}" -f bmp -o symbol-textrefine -F jb2 -ini \
jbig2-symbol-textrefine.ini
Template 2 is needed by some symbols in 0000372.pdf page 11 and
0000857.pdf pages 1-4. Implement the others too while here. (The
mentioned pages in those two PDFs also use the "end of stripe" segment,
so they still don't render yet.
We still don't support EXTTEMPLATE.
This extracts the bitbuffer combining code we had into a new function
composite_bitbuffer() and adds the following features:
* Real support for combination operators (which also lets us allow black
as background color again, even if that's never used in practice)
* Clipping support (not used here yet, but will be needed elsewhere
soon)
We're going to need this for text segment handling.
No behavior change.
I created this file using `jbig2` (see below for details), but as
far as I can tell `jbig2` does not produce spec-compliant files:
1. It always writes to 0s for the run lengths that specify how
many symbols to export at the end of a symbol segment
2. It doesn't write any referred-to segments for text segments.
I think it's supposed to write a referred-to segment that
mentions the symbol segment the text segment refers to (?)
I locally tweaked `jbig2` to fix these two defects (*), so the image
added in this commit is correct as best I can tell. It opens fine
using `image` and `jbig2`'s decode mode, and via
`Meta/jbig2_to_pdf.py` in Firefox and Chrome. Without my tweaks,
the image decodes fine with `jbig2`, but not with any of the other
three. The image (in a pdf) does _not_ decode in Preview.app,
either with or without my local `jbig2` tweaks.
*: See the PR adding this image for my local diff.
I created the test image file by running this shell script with
`jbig2` tweaked as described above:
#!/bin/bash
set -eu
I=Build/lagom/bin/image
S=Tests/LibGfx/test-inputs/bmp/bitmap.bmp
$I "$S" --crop 232,70,120,250 -o mouth.bmp
$I "$S" --crop 135,100,100,100 -o nose.bmp
$I "$S" --crop 50,108,30,30 -o top_eye.bmp
$I "$S" --crop 60,265,30,30 -o bottom_eye.bmp
# I then manually converted those to 1bpp using Photoshop
# (Image->Mode->Grayscale, then Image->Mode->Bitmap...,
# File->Save As..., bmp) since `jbig2` gets confused by non-1bpp
# bmp files and `image` can't write 1bpp files :/
#
# (I tried `convert ${in} -monochrome ${in}-1bpp.bmp` via
# https://cancerberosgx.github.io/magic/playground/index.html
# first, but that produced bmp files that neither Preview.app nor
# `jbig2` could handle.)
#
# -HeightClass: Number of height classes
# -WidthClass: Maximum number of symbols in one height class
# -Simple means no refinement; the number after is the symbol's ID
# The 3 numbers afer `-ID` are id, y, x. The `-ID` are sorted by x.
# -RefCorner 1 means "top left".
#
# `jbig2` writes symbol and text segments as specified in the ini
# file, and then only stores the bits of the input image that aren't
# already set through symbol and text segments.
cat << EOF > jbig2-symbol.ini
-sym -Seg 1
-sym -file -numClass -HeightClass 3 -WidthClass 2
-sym -file -numSymbol 4
-sym -file -Height 250
-sym -file -Width 120 -Simple 0 mouth-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -file -Height 100
-sym -file -Width 100 -Simple 1 nose-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -file -Height 30
-sym -file -Width 30 -Simple 2 top_eye-1bpp.bmp
-sym -file -Width 30 -Simple 3 bottom_eye-1bpp.bmp
-sym -file -EndOfHeightClass
-sym -Param -Huff_DH 0
-sym -Param -Huff_DW 0
-txt -Seg 2
-txt -Param -numInst 4
-ID 2 108 50 -ID 3 265 60 -ID 1 100 135 -ID 0 70 232
-txt -Param -RefCorner 1
-txt -Param -Xlocation 0
-txt -Param -Ylocation 0
-txt -Param -W 399
-txt -Param -H 400
EOF
J=$HOME/Downloads/T-REC-T.88-201808-I\!\!SOFT-ZST-E/Software
J=$J/JBIG2_SampleSoftware-A20180829/source/jbig2
$J -i "${S%.bmp}" -f bmp -o symbol -F jb2 -ini jbig2-symbol.ini
"TPGD" is short for "Typical Prediction for Generic Direct coding",
and the "ON" bit turns it on. In this mode, before decoding a line,
we decode a single bit first that controls if the current line is
just a copy of the previous line. If so, the line's pixels aren't
encoded, the decoder just copies the previous line.
I created this by running
jbig2 -i Tests/LibGfx/test-inputs/bmp/bitmap -f bmp \
-o bitmap -F jb2 -ini tpgdon.ini
where tpgdon.ini contained:
-Gen -Seg 1
-Gen -Param -TpGDon 1
See previous commits in this directory for details on the `jbig2` tool.
Sadly, the TPGDON writing path in `jbig2` wasn't implemented yet,
so I had to add this. See the PR that added this commit for my
local diff to `jbig2`.
I'm somewhat confident that my change to `jbig2` (and hence the
image added in this commit) is correct because:
1. `jbig2` succeeds in converting this file to a bmp file,
while it failed without my patch (the decoding codepath in
`jbig2` does have TPGDON support)
2. Other pdf viewers display the output of
`Meta/jbig2_to_pdf.py -o foo.pdf path/to/bitmap-tpgdon.jbig2 399 400`
the same way we do
It seems to do the right thing already, and nothing in the spec says
not to do this as far as I can tell.
With this, we can finally decode
Tests/LibGfx/test-inputs/jbig2/bitmap.jbig2 and add a test for
decoding simple arithmetic-coded images.
In practice, everything uses white backgrounds and operators `or`
or `xor` to turn them black, at least for the simple images we're
about to be able to decode.
To make sure we don't forget implementing this for real once needed,
reject other ops, and also reject black backgrounds (because 1 | 0
is 1, not 0 like our overwrite implementation will produce).
This means we have to remove a test, but since this scenario doesn't
seem to happen in practice, that seems ok.
The context can vary for every bit we read.
This does not affect the one use in the test which reuses the same
context for all bits, but it is necessary for future changes.
I think the context normally changes for every bit. But this here
is enough to correctly decode the test bitstream in Annex H.2 in
the spec, which seems like a good checkpoint.
The internals of the decoder use spec naming, to make the code
look virtually identical to what's in the spec. (Even so, I managed
to put in several typos that took a while to track down.)
With this, `image` can convert any jbig2 file, as long as it's
black (or white), and LibPDF can draw jbig2 files (again, as long
as they only contain a single color stored in just a
PageInformation segment).
I extracted a pure-white jbig2 that has only a PageInformation segment
from a PDF and manually edited the bytes to reduce the bitmap size to
47x23 and to clear all unneded bits (except, in the black version,
the page color bit is set).
This allows `file` to correctly print the dimensions of a .jbig2 file,
and it allows us to write a test that covers much of all the code
written so far.
I created this by running
jbig2 -i Tests/LibGfx/test-inputs/bmp/bitmap -f bmp -o bitmap -F jb2
using the `jbig2` tool whose source code is in the zip file here:
https://www.itu.int/rec/T-REC-T.88-201808-I
(Just `make jbig2` in Software/JBIG2_SampleSoftware-A20180829/source
was enough to build it.)
As far as I can tell (cf `JBIG2_EncMain()` which always writes `1`
to the flags byte in the file header), this tool always writes files
in the sequential organization.
This is for validating that a decoder with a weak or nonexistent
sniff() method thinks it can decode an image. This should not be
treated as an error.
No behavior change.
The semantics of BGRx8888 aren't super clear and it means different
things for different parts of the codebase. In particular, the PNG
writer still writes the x channel to the alpha channel of its output.
In BMPs, the 4th palette byte is usually 0, which means after #21412 we
started writing all .bmp files with <= 8bpp as completely transparent
to PNGs.
This works around that.
(See also #19464 for previous similar workarounds.)
The added `bitmap.bmp` is a 1bpp file I drew in Photoshop and saved
using its "Save as..." saving path.
A tile is basically a strip with a user-defined width. With that in
mind, adding support for them is quite straightforward. As a lot the
common code was named after 'strips', to avoid future confusion I
renamed everything that interact with either strips or tiles to a
global term: 'segment'.
Note that tiled images are supposed to always have a 'TileOffsets' tag
instead of 'StripOffset'. However, this doesn't seem to be enforced by
encoders, so we support having either of them indifferently.
The test case was generated with the following Python script:
import pyvips
img = pyvips.Image.new_from_file('deflate.tiff')
img.write_to_file('tiled.tiff',
compression=pyvips.ForeignTiffCompression.DEFLATE,
tile=True, tile_width=64, tile_height=64)
JPEGs can store a `restart_interval`, which controls how many
minimum coded units (MCUs) apart the stream state resets.
This can be used for error correction, decoding parts of a jpeg
in parallel, etc.
We tried to use
u32 i = vcursor * context.mblock_meta.hpadded_count + hcursor;
i % (context.dc_restart_interval *
context.sampling_factors.vertical *
context.sampling_factors.horizontal) == 0
to check if we hit a multiple of an MCU.
`hcursor` is the horizontal offset into 8x8 blocks, vcursor the
vertical offset, and hpadded_count stores how many 8x8 blocks
we have per row, padded to a multiple of the sampling factor.
This isn't quite right if hcursor isn't divisible by both
the vertical and horizontal sampling factor. Tweak things so
that they work.
Also rename `i` to `number_of_mcus_decoded_so_far` since that
what it is, at least now.
For the test case, I converted an existing image to a ppm:
Build/lagom/bin/image -o out.ppm \
Tests/LibGfx/test-inputs/jpg/12-bit.jpg
Then I resized it to 102x77px in Photoshop and saved it again.
Then I turned it into a jpeg like so:
path/to/cjpeg \
-outfile Tests/LibGfx/test-inputs/jpg/odd-restart.jpg \
-sample 2x2,1x1,1x1 -quality 5 -restart 3B out.ppm
The trick here is to:
a) Pick a size that's not divisible by the data size width (8),
and that when rounded to a block size (13) still isn't divisible
by the subsample factor -- done by picking a width of 102.
b) Pick a huffman table that doesn't happen to contain the bit
pattern for a restart marker, so that reading a restart marker
from the bitstream as data causes a failure (-quality 5 happens
to do this)
c) Pick a restart interval where we fail to skip it if our calculation
is off (-restart 3B)
Together with #22987, fixes#22780.
Non-interleaved files always have an MCU of one data unit.
(A "data unit" is an 8x8 tile of pixels, and an "MCU" is a
"minium coded unit", e.g. 2x2 data units for luminance and
1 data unit each for Cr and Cb for a YCrCb image with
4:2:0 subsampling.)
For the test case, I converted an existing image to a ppm:
Build/lagom/bin/image -o out.ppm \
Tests/LibGfx/test-inputs/jpg/12-bit.jpg
Then I converted it to grayscale and saved it as a pgm in Photoshop.
Then I turned it into a weird jpeg like so:
path/to/cjpeg \
-outfile Tests/LibGfx/test-inputs/jpg/grayscale_mcu.jpg \
-sample 2x2 -restart 3 out.pgm
Makes 3 of the 5 jpegs failing to decode at #22780 go.
I opened Base/res/graphics/buggie.png in Photoshop, converted it
to U.S. Web Coated (SWOP) v2, flattened the image so we don't have
CMYK with alpha, and saved it as a jpeg (with color profile embedded).
When present, the alpha channel is also affected by the horizontal
differencing predictor.
The test case was generated with GIMP with the following steps:
- Open an RGB image
- Add a transparency layer
- Export as TIFF with the LZW compression scheme
This tag is required by the specification, but some encoders (at least
Krita) don't write it for images with a single strip.
The test file was generated by opening deflate.tiff in Krita and saving
it with the DEFLATE compression.
Type 2 <=> One-dimensional Group3, customized for TIFF
Type 3 <=> Two-dimensional Group3, uses the original 1D internally
Type 4 <=> Two-dimensional Group4
So let's clarify that this is not Group3 1D but the TIFF variant, which
is called `CCITTRLE` in libtiff. So let's stick with this name to avoid
confusion.
Images with a display mask ("stencil" as it's called in DPaint) add
an extra bitplane which acts as a mask. For now, at least skip it
properly. Later we should render masked pixels as transparent, but
this requires some refactoring.
