-> split libraries further into decoder / encoder
-> add libwebpdecoder.a in Makefile.unix
-> make dwebp link against libwebpdecoder.a in Makefile.unix
also: in makefile.unix, pass EXTRA_FLAGS to LDFLAGS too
(otherwise, -m32 wouldn't work, e.g.)
Change-Id: Ief3da02a729dd86bbaf949ed048836716941657f
Simplify and re-organize the VP8L bit-reader functions
(e.g.: the 40-bit look-ahead code was helping much)
Speed-up with LBITS=64, on arm7-a:
=> before:
./dwebp_justify_24_neon -v bryce_ll.webp
Time to decode picture: 11.393s
File bryce_ll.webp can be decoded (dimensions: 11158 x 2156).
...
=> after (LBITS=64): Time to decode picture: 9.953s
making the VP8L bit-reader in 32 bit mode is going to be
harder (because we need to be able to read two symbols
at a time, each with max length 15 bits)
Change-Id: I89746fb103b87b5e2fd40a3208a6fbc584b88297
If a NULL pre-allocated buffer is passed, a buffer will be automatically
allocated.
+ add some parameter checks.
reported in http://code.google.com/p/webp/issues/detail?id=139
Change-Id: I9e14ed97db30ee12e46b5e92aac7eeaaeb99bfd5
- precompute filtering strength once for all at the beginning
instead of per-macroblock
- reduce size of VP8MB struct from 8 bytes to 4.
- removed VP8StoreBlock() accordingly
Change-Id: Icf3d329473e21c464770be3d72a04c9ee4c321f2
GetCoeffs is (by far) the most consuming function of the decoder.
No speed change (unfortunately), but the main loop is somehow clearer.
Change-Id: I78f1c10cadc2c8696c041f5cbda86cab92cc6598
- Separate out mux.h and demux.h
- muxtypes.h: new header for data types common to mux/demux
- Move some misc read/write utilities to utils/utils.h
- Remove some duplicate methods.
- Separate out mux/demux libraries
Change-Id: If9b9569b10d55d922ad9317ef51710544315d6de
- Also, use the term 'fragments' instead of 'tiling' in code
- This makes code consistent with the spec.
Change-Id: Ibeccffc35db23bbedb88cc5e18e29e51621931f8
Use separate fourCCs "XMP " and "EXIF" instead of a common "META"
Also, some refactorization in webpmux.c
Change-Id: Iad3337e5c1b81e785c60670ce28b1f536dd7ee31
the max wasn't checked leading to a rollover case, possibly exploitable.
additionally check the RIFF size early, to avoid similar issues.
pulled from chromium:
http://codereview.chromium.org/11229048/
Change-Id: I4050b13a7e61ec023c0ef50958c45f651cf34c49
lossy was rounding with a bias toward opaque:
[232+, 8] -> [15, 1]
now both paths use the range:
[240+, 16] -> [15, 1]
Change-Id: I3da2063b4959b9e9f45bae09e640acc1f43470c5
With this, MODE_rgbA can safely be used without speed penalty
even in case of pure-lossy alpha-less input.
It's also an optimization when cropping a fully-opaque region from
an image with alpha: premultiply is then skipped
Change-Id: Ibee28c75744f193dacdfccd5a2e7cd1e44604db6
Fix the lossless decoder for the case when it has to apply other
inverse transforms before applying Color indexing inverse transform.
The main idea is to make ColorIndexingInverse virtually in-place: we
use the fact that the argb_cache is allocated to accommodate all
*unpacked* pixels of a macro-row, not just *packed* pixels.
Change-Id: I27f11f3043f863dfd753cc2580bc5b36376800c4
minor revision shouldn't matter, we only check major revision number.
Bumped all version numbers so that incompatibility starts *now*
Change-Id: Id06c20f03039845ae4cfb3fd121807b931d67ee4
the VP8 decode functions do not need to be public; only
GetInfo/CheckSignature need to be marked extern as they're used by
libwebpmux.
Change-Id: Id9ab4d6166b0271cf5d04563c6dac1fcc84adbdc
could have led to some negative overflow on 32bit arch
(if it was not for the "total_size == (size_t)total_size" test)
Change-Id: I7640340b605b9c674d30dd58a1e2144707299683
Check for valid bounds on the 'dist' in backward reference case.
Clamp it to 1 in case of zero and negative values.
Change-Id: I78e956d4595955efa02b1f9628b475093f6ee001
moves the implementation to ParseHeadersInternal. this also allows
decoding to start at a VP8X sub-chunk, e.g. 'ALPH'.
Change-Id: I06791f87d90f888de32746ecb02705e4b0ff227a
ParseVP8X was checking for presence of extra 20 bytes (after RIFF header).
This check should not be executed for non-mux (non-VP8X) images.
Change-Id: I3fc89fa098ac0a53102e7bbf6c291269817c8e47
Ensure that the lossless bit-stream doesn't allow for such cases and
safe-gaurd decoder against indefinite recursion.
Change-Id: Ia6d7f519291de8739f79a977a5800982872aae71
The new modes are
MODE_rgbA
MODE_bgrA
MODE_Argb
MODE_rgbA_4444
It's binary incompatible, since the enums changed.
While at it, i removed the now unneeded KeepAlpha methods.
-> Saved ~12k of code!
* made explicit mention that alpha_plane is persistent,
so we have access to the full alpha plane data at all time.
Incremental decoding of alpha was planned for, but not
implemented. So better not dragged this constaint for now
and make the code easier until we revisit that.
Change-Id: Idaba281a6ca819965ca062d1c23329f36d90c7ff
Change the lossless signature to 0x2f
Add 1 bit indicator for 'droppable (or trivial) alpha)'.
Add 3 bit lossless version (for future extension like yuv support).
Change the sub-resolution information to 3 bits implying range [2 .. 9]
Change-Id: Ic7b8c069240bbcd326cf5d5d4cd2dde8667851e2
Previously, it used to assume any raw bitstream is a VP8 one.
Also,
- Factor out VP8CheckSignature() & VP8LCheckSignature().
- Use a local var for *data_ptr in ParseVP8Header() for
readability.
Change-Id: I0fa8aa177dad7865e00c8898f7e7ce76a9db19d5
frame's InitIo should not reset fancy_upsampling flag.
This flag (fancy_upsampling) is set via CustomSetup -> WebPIoInitFromOptions
and frame's InitIo() is resetting it to 0.
Change-Id: I64b54cdfba43799c0a5aa8e384575af5d6331674
may be useful later for instance to bypass some code
if we know we don't use the Bundle+ColorMap transform.
Change-Id: I9dc70d18165b2363ad9ede763684ef3d8eba5903
This saves ~26 bytes of headers.
* introduce new VP8LDecodeAlphaImageStream() for decoding
* use VP8LEncodeStream() for encoding
* refactor code a bit
still TODO: make the alpha-quality/enc-method user-configurable
Change-Id: I23e599bebe335cfb5868e746e076c3358ef12e71
Limit the overall number of transformations to 4 and disallow any
duplicate transform for decoding an image.
Change-Id: Ic4b0ecd553db96702e117fd073617237d95e45c0
this allows later customization of data output method.
No perf diff observed, even if ProcessRows is no longer inlined.
Change-Id: I6933a3612a9cf6c108cf2776dfde0ae80c6c07c0
+ small opportunistic fixes:
* allow NULL decoded_data to be passed to DecodeStream
and clarity (with assert()) when to do so
* AllocateAndInitRescaler() was already setting error status,
as it should. No need to do it at caller's site
Change-Id: I30867e596564a7f459a0d1ddbf6f5d312414b7fd
now, we only use 2 bits for the filtering method, and 2 bits
for the compression method.
There's two additional bits which are INFORMATIVE, to specify
whether the source has been pre-processed (level reduction)
during compression. This can be used at decompression time
for some post-processing (see DequantizeLevels()).
New relevant spec excerpt:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| ChunkHeader('ALPH') |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|Rsv| P | F | C | Alpha Bitstream... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Compression method (C): 2 bits
: The compression method used:
* `0`: No compression.
* `1`: Backward reference counts encoded with arithmetic encoder.
Filtering method (F): 2 bits
: The filtering method used:
* `0`: None.
* `1`: Horizontal filter.
* `2`: Vertical filter.
* `3`: Gradient filter.
Pre-processing (P): 2 bits
: These INFORMATIVE bits are used to signal the pre-processing that has
been performed during compression. The decoder can use this information to
e.g. dither the values or smooth the gradients prior to display.
* `0`: no pre-processing
* `1`: level reduction
Decoders are not required to use this information in any specified way.
Reserved (Rsv): 2 bits
: SHOULD be `0`.
Alpha bitstream: _Chunk Size_ - `1` bytes
: Encoded alpha bitstream.
This optional chunk contains encoded alpha data for a single tile.
Either **ALL or NONE** of the tiles must contain this chunk.
The alpha channel data is losslessly stored as raw data (when
compression method is '0') or compressed using the lossless format
(when the compression method is '1').
Change-Id: Ied8f5fb922707a953e6a2b601c69c73e552dda6b
* Method #1 is now calling the lossless encoder on the alpha plane.
Format is not final, it's just a first draft. We need ad-hoc functions.
* removed now useless utils/alpha.*
* added utils/quant_levels.h instead
* removed the TCoder code altogether
Change-Id: I636840b6129a43171b74860e0a0fc5bb1bcffc6a
When we are at end-of-stream, but haven't decoded all pixels, we should
return an error.
Also remove an obsolete TODO.
Change-Id: I3fb1646136e706da536d537a54d1fa487a890630
- Symbols added to the tree are valid inside HuffmanTreeBuildExplicit().
- In HuffmanTreeBuildImplicit(), make sure 'root_symbol' is
valid in case of a single symbol tree.
Change-Id: I7de5de71ff28f41e2d6228b29ed8dd4a20813e99
[Basically, the condition "src - dist < data" can be wrongly evaluated
to be false if "src < dist" due to underflow. Instead, "src - data <
dist" is the correct condition, as "src > data" is always true and so
there would never be an underflow].
Change-Id: Ic9f64bfe76a9acae97abc1fb7c1f4868e81f1eb8
No empty trees are codified with the simple Huffman code. The simple Huffman
code is simplified to be either a 1-bit code or 8-bit code for symbols.
Change-Id: I3e2813027b5a643862729339303d80197c497aff
* don't transmit the number of Huffman tree group explicitly
* move color-cache information before the meta-Huffman block
* also add a check that color_cache_bits is in [1..11] range, as per spec.
Change-Id: I81d7711068653b509cdbc1151d93e229c4254580
src/dec/Makefile.am: add missing reference to vp8li.h
src/{dec,dsp,enc}/Makefile.am: move some headers to noinst_
Change-Id: I0e2bc69980bd8175d99ad0ab63f537ef9e425b77
- use common file organization across subdir makefiles
- append lib/source/header list variables and sort
Change-Id: I0653e1c73a4552b0c43d21f321b22b4972d6e87b
* each with their own decoder instances.
* Refactor the incremental buffer-update code a lot.
* remove br_offset_ for VP8LDecoder along the way
* make VP8GetHeaders() be used only for VP8, not VP8L bitstream
* remove VP8LInitDecoder()
* rename VP8LBitReaderResize() to VP8LBitReaderSetBuffer()
(cherry picked from commit 5529a2e6d47212a721ca4ab003215f97bd88ebb4)
Change-Id: I58f0b8abe1ef31c8b0e1a6175d2d86b863793ead
import changes from experimental 5529a2e^
and enable build in autoconf and makefile.unix; windows will be treated
separately.
Change-Id: Ie2e177a99db63190b4cd647b3edee3b4e13719e9
Pulled from 5529a2e^.
The history of this and related files is a bit entangled so rather
trying to split the changes and introduce some noise in master's history
we'll start with a fresh snapshot.
The file progression is still available in the experimental branch.
Change-Id: I490a49cb0084d556a29d7c5257c986ea695ce60c
A call to Append/Update would index the parts_ array w/-1 as num_parts_
had yet to be set by DecodePartition0. This would cause corruption
within the VP8Decoder member.
Fixes issue #106.
Change-Id: Ib9f2811594ff19e948a66fda862a4e0a384bb9aa
If the first call to WebPIUpdate contained all of partition 0, but not
enough to decode a macroblock the test for max macro block size could
fail (4096 bytes) due to a memory buffer offset not being updated.
This change offsets the buffer in the same manner as Append().
Fixes issue #105.
Change-Id: I90ca3d0dde92eedc076933b8349d2d297f2469e8
- remove some unused functions
- move global arrays from data to read only section
- explicitly cast malloc returns; not specifically necessary, but helps
show intent
- miscellaneous formatting
Change-Id: Ib15fe5b37fe6c29c369ad928bdc3a7290cd13c84
Fix bug for Alpha data in RGBA_4444 color-mode.
The Alpha data is required to be clipped [0, 15] while
converting from 8 bits to 4 bits.
Change-Id: I80705d575c35121beb9633a05ec8823435c79586
alpha information is not to be found at RIFF chunks level, not
in the VP8 bitstream (that was a tmp hack)
Change-Id: Idd1629c696b03c26f6f30650d7216f627f1761df
- Fix the off-by-one diff when cropping with simple-filter.
- Fix a bug in incremental decoding in case of alpha.
- In VP8FinishRow(), do not decode alpha when y_start > y_end.
- Correct output of alpha channel for MODE_ARGB.
- Correct output of alpha channel for MODE_RGBA_4444.
Change-Id: I785763a2a704b973cc742ad93ffbb53699d1fc0a
Extend WebP decode functionality to extract Alpha data (support ALPH
chunk) and decode the Alpha plane using Alpha (utils/alpha) core-lib.
Change-Id: I6f0ee1b189c13caba3c1dd9b681383bfd00aa212
For odd-sized chunk-payload, there's one Byte padding at the end to make the
next chunk start at even-address. The chunk-skip logic needs to be updated to
handle such cases.
Change-Id: I8939ed07584a6195092fbe703eb0ea711fa5f153
- Move common defines to dec/webpi.h
- Regularize naming and parameters of various "CheckAndSkip" functions.
Also they return VP8StatusCode for clarity.
- Move WebP header/chunk parsing functions to webpi.h
- Fix a bug in static function GetFeatures()
Change-Id: Ibc3da4ba210d860551724dc40741959750d66b89
Gathers all DSP-related function (and SSE2 implementations).
Clean-up some unwanted symbolic dependencies so that webp_encode,
webp_decode and webp_dsp are truly independent libraries.
+ opportunistic clean-up:
* remove unneeded VP8DspInitTables(), now integrated in VP8DspInit()
* make consistent use of VP8GetCPUInfo() in the various DspInit() funcs
* change OUT macro to DST
with assertions enabled the code would abort in
WebPWorkerChangeState with:
Assertion `worker->status_ >= OK'
without them the code would hang in the _cond_wait.
this change makes WebPWorkerChangeState a no-op in this case.
Change-Id: Iea855568bbdef2865ae61ab54473b3a7c230e91a
To be enabled with the flag WEBP_USE_THREAD.
For now it's only available on unix (pthread), when using Makefile.unix
Will be switched on more generally later.
In-loop filtering and output (=rescaling/yuv->rgb conversion)
is done in parallel to bitstream decoding, lagging 1 row behind.
Example:
examples/dwebp bryce.webp -v
Time to decode picture: 0.680s
examples/dwebp bryce.webp -v -mt
Time to decode picture: 0.515s
Change-Id: Ic30a897423137a3bdace9c4e30465ef758fe53f2
RGB565 and ARGB4444 are only supported through the advanced decoding API.
ARGB being somewhat generic, there's an easy WebPDecodeARGB()
new function for convenience.
Patch by Vikas Arora (vikaas dot arora at gmail dot com)
Change-Id: Ic7b6f72bd70aca458d14e7fdd23679212430ebca
~5-10% faster.
Heavy 8bit arithmetic trickery!
Patch by Somnath Banerjee (somnath at google dot com)
Change-Id: I9fd2c511d9f631e9cf4b008c46127b49fb527b47
You can now use WebPDecBuffer, WebPBitstreamFeatures and WebPDecoderOptions
to have better control over the decoding process (and the speed/quality tradeoff).
WebPDecoderOptions allow to:
- turn fancy upsampler on/off
- turn in-loop filter on/off
- perform on-the-fly cropping
- perform on the-fly rescale
(and more to come. Not all features are implemented yet).
On-the-fly cropping and scaling allow to save quite some memory
(as the decoding operation will now scale with the output's size, not
the input's one). It saves some CPU too (since for instance,
in-loop filtering is partially turned off where it doesn't matter,
and some YUV->RGB conversion operations are ommitted too).
The scaler uses summed area, so is mainly meant to be used for
downscaling (like: for generating thumbnails or previews).
Incremental decoding works with these new options.
More doc to come soon.
dwebp is now using the new decoding interface, with the new flags:
-nofancy
-nofilter
-crop top left width height
-scale width height
Change-Id: I08baf2fa291941686f4ef70a9cc2e4137874e85e
This is a temporary workaround for issue #80.
Currently dec/dsp_sse2 is not included in Makefile.vc. Simply adding it
will cause a build error, however, as cl does not support aligning
function parameters producing, e.g.,
src\dec\dsp_sse2.c(228) : error C2719: 'q1': formal parameter with
__declspec(align('16')) won't be aligned
Change-Id: Id29e6802dd29110e59c4f6d13ffa5d4793c750a0
There was 1 unneeded sample line allocated for the filter cache in of simple filtering.
+ Add an explaining comment.
Change-Id: I775a596c8b8643e773e0eade8aa341dc23fb290f
+ add a simple rescaling function: WebPPictureRescale() for encoding
+ clean-up the memory managment around the alpha plane
+ fix some includes path by using "../webp/xxx.h" instead of "webp/xxx.h"
New flags for 'cwebp':
-resize <width> <height>
-444 (no effect)
-422 (no effect)
-400
Change-Id: I25a95f901493f939c2dd789e658493b83bd1abfa
This is a (minor) bitstream change: if the 'color_space' bit is set to '1'
(which is normally an undefined/invalid behaviour), we add extra data at the
end of partition #0 (so-called 'extensions')
Namely, we add the size of the extension data as 3 bytes (little-endian),
followed by a set of bits telling which extensions we're incorporating.
The data then _preceeds_ this trailing tags.
This is all experimental, and you'll need to have
'#define WEBP_EXPERIMENTAL_FEATURES' in webp/types.h to enable this code
(at your own risk! :))
Still, this hack produces almost-valid WebP file for decoders that don't
check this color_space bit. In particular, previous 'dwebp' (and for instance
Chrome) will recognize this files and decode them, but without the alpha
of course. Other decoder will just see random extra stuff at the end of
partition #0.
To experiment with the alpha-channel, you need to compile on Unix platform
and use PNGs for input/output.
If 'alpha.png' is a source with alpha channel, then you can try (on Unix):
cwebp alpha.png -o alpha.webp
dwebp alpha.webp -o test.png
cwebp now has a '-noalpha' flag to ignore any alpha information from the
source, if present.
More hacking and experimenting welcome!
Change-Id: I3c7b1fd8411c9e7a9f77690e898479ad85c52f3e
Makes things lighter at the expense of requiring the user
to be up-to-date for autotools.
patch by Jan Engelhardt (jengelh at medozas dot de)
Change-Id: Icfcab2d899828a213d9fade0dab350dacd0c070a
going down to strict -ansi c89 is quite overkill (no 'inline',
and /* */-style comments).
But with these fixes, the code compiles with the stringent flags:
-Wextra -Wold-style-definition -Wmissing-prototypes
-Wmissing-declarations and -Wdeclaration-after-statement
Change-Id: I36222f8f505bcba3d9d1309ad98b5ccb04ec17e3
WebPGetDecoderVersion() and WebPGetEncoderVersion()
will not return 0.1.2 encoded as 0x000102
dwebp and cwebp also have a new "-version" flag
Change-Id: I4fb4b5a8fc4e53681a386ff4b74fffb639fa237a
The object WebPIDecoder is available to store the
decoding state. The flow is typically:
WebPIDecoder* const idec = WebPINew(mode);
while (has_more_data) {
// ... (get additional data)
status = WebPIAppend(idec, new_data, new_data_size);
if (status != VP8_STATUS_SUSPENDED ||
break;
}
// The above call decodes the current available buffer.
// Part of the image can now be refreshed by calling to
// WebPIDecGetRGB()/WebPIDecGetYUV() etc.
}
WebPIDelete(idec);
Doing so, one can try and decode new macroblocks everytime fresh
bytes are available.
There's two operating modes: either appending fresh bytes, or
updating the whole buffer with additional data in the end.
The latter requires less memcpy()'s
main patch by Somnath Banerjee (somnath at google.com)
Change-Id: Ie81cbd0b50f175743af06b1f964de838b9a10a4a
use top_srcdir rather than top_builddir for AM_CPPFLAGS
add EXTRA_DIST to man Makefile. fixes distcheck target.
Change-Id: I308dc1c98f096de1efe188f63d040ef953598e78
non-zero dc-bit was mixed with non-zero ac-bit, preventing
finer optimization during VP8ReconstructBlock.
Depending on sparsity, i see 2-5% gain on average.
Change-Id: I7f34f18d0701c77837de3540b732e5b7d85d7c5d
(this makes initialization easier and will be helpful for incremental
decoding).
Modify ParsePartitions() to accommodate for truncated input.
Change-Id: I62f52078d6b7a2314a11880a20d9eac5b4714bd0
libvpx (and ffvp8) implementations are completely skipping the deblocking
step if loop_filter_level is 0, which is _not_ equivalent to performing
the loop-filtering with a 0 value for loop_filter_level. In the latter case
(which we followed), few pixels were modified here and there and you could
observe off-by-1 errors on few places.
This patch will reconcile the 3 implementations (since the difference
is minor, skipping the deblocking step will save CPU for virtually
no visible difference).
The spec will be made clearer about the expected behaviour:
* if the global loop_filter_level is 0, turn deblocking off.
* if it's not 0 but the local loop_filter_level ends up being 0 for whatever
reason (lf_delta, mode delta, ref delta, etc.) on a particular
macroblock, skip the deblocking too.
Change-Id: I157f1f8de463b8a76caddb3f347b7fbc7bd527d2
This will make the decoder skip the filtering process if needed,
resulting in speed-up, but also non-compliant (blocky?) output
+ Add a versioning check for VP8InitIo(), since we've adding a field to VP8Io
+ add some more error checks while at it
Change-Id: I4e9899edc24ecf8600cbb27aa4038490b7b2cef3
converts PNG & JPEG to WebP
This is an experimental early version, with lot of room
of later optimizations in both speed and quality.
Compile with the usual `./configure && make`
Command line example is examples/cwebp
Usage:
cwebp [options] -q quality input.png -o output.webp
where 'quality' is between 0 (poor) to 100 (very good).
Typical value is around 80.
More encoding options with 'cwebp -longhelp'
Change-Id: I577a94f6f622a0c44bdfa9daf1086ace89d45539
For very very short partitions, the initial GetByte() could
set eof_ to 1, whereas some bits were available but unread yet.
So we set eof_ to 0 last.
Change-Id: Ic6b68271bc72efa4de4e64e1f57307d1d8fb613c
