Changelog
All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
The changes are relative to the previous release, unless the baseline is specified.
Added since 1.1.1
- Add the properties and numProperties fields to avifImage. They are filled by the avifDecoder instance with the properties unrecognized by libavif.
Changed since 1.1.1
- avifenc: Allow large images to be encoded.
- Fix empty CMAKE_CXX_FLAGS_RELEASE if -DAVIF_CODEC_AOM=LOCAL -DAVIF_LIBYUV=OFF is specified. https://github.com/AOMediaCodec/libavif/issues/2365.
- Renamed AVIF_ENABLE_EXPERIMENTAL_METAV1 to AVIF_ENABLE_EXPERIMENTAL_MINI and updated the experimental reduced header feature to the latest specification draft.
- Ignore gain maps with unsupported metadata. Handle gain maps with writer_version > 0 correctly. Simplify gain map API: remove the enableParsingGainMapMetadata setting, now gain map metadata is always parsed if present and if this feature is compiled in. Replace enableDecodingGainMap and ignoreColorAndAlpha with a bit field to choose image content to decode. Remove gainMapPresent: users can check if decoder->image->gainMap != NULL instead.
- Write an empty HandlerBox name field instead of “libavif” (saves 7 bytes).
- Update aom.cmd/LocalAom.cmake: v3.11.0
- Update avm.cmd: research-v8.1.0
- Update dav1d.cmd/dav1d_android.sh/LocalDav1d.cmake: 1.5.0
- Update libjpeg.cmd/LocalJpeg.cmake: v3.0.4
- Update libxml2.cmd/LocalLibXml2.cmake: v2.13.4
- Update svt.cmd/svt.sh/LocalSvt.cmake: v2.2.1
- Change experimental gainmap API: remove avifGainMapMetadata and avifGainMapMetadataDouble structs.
- Add avif(Un)SignedFraction structs and avifDoubleTo(Un)SignedFraction utility functions.
- Turn on the gain map API by default. Rename the compile flag from AVIF_ENABLE_EXPERIMENTAL_GAIN_MAP to AVIF_ENABLE_GAIN_MAP.
1.1.1 - 2024-07-30
Changed since 1.1.0
- In avif.h, change “AVIF_API AVIF_NODISCARD” back to “AVIF_NODISCARD AVIF_API” to fix clang-cl and MSVC compilation errors in the shared library build on Windows.
- Fix -DAVIF_GTEST=SYSTEM, https://github.com/AOMediaCodec/libavif/issues/2258.
- Fix infe_type and codec_config_type wrongly read as byte-aligned fields in the experimental feature AVIF_ENABLE_EXPERIMENTAL_METAV1.
- When building aom as a local dependency, runtime CPU detection (
CONFIG_RUNTIME_CPU_DETECT
) is now always ON
; in 1.1.0 it had been disabled for non-native builds. - Fix CMake config shared library leaks https://github.com/AOMediaCodec/libavif/issues/2264.
- Fix clang-cl compilation.
- Update gain map metadata to current ISO 21496-1 draft.
- cmake: Only search for ASM_NASM language on x86_64 platforms.
- Fix “No known features for CXX compiler” CMake error.
- Fix aom link flags so that transitive library link flags are included when aom is a static library https://github.com/AOMediaCodec/libavif/issues/2274.
- Fix out-of-order ‘dimg’ grid associations https://github.com/AOMediaCodec/libavif/issues/2311.
- Report files with an item used in multiple ‘dimg’ boxes with AVIF_RESULT_NOT_IMPLEMENTED instead of AVIF_RESULT_INVALID_IMAGE_GRID.
1.1.0 - 2024-07-11
Added since 1.0.0
- Add experimental API for reading and writing gain maps in AVIF files. If enabled at compile time, add
gainMap
field to avifImage
, add qualityGainMap
field to avifEncoder
, add gainMapPresent
, enableDecodingGainMap
, enableParsingGainMapMetadata
and ignoreColorAndAlpha
to avifDecoder
. Utility functions for working with gain maps are also added. Gain maps allow readers that support them to display HDR images that look good on both HDR and SDR displays. This feature is highly experimental. The API might change or be removed in the future. Files created now might not decode in a future version. This feature is off by default and must be enabled with the AVIF_ENABLE_EXPERIMENTAL_GAIN_MAP compilation flag. - Add experimental support for converting jpeg files with gain maps to AVIF files with gain maps. Requires libxml2, and the AVIF_ENABLE_EXPERIMENTAL_GAIN_MAP compilation flag. Add a --qgain-map flag to control the gain map quality in avifenc.
- Add the headerFormat member of new type avifHeaderFormat to avifEncoder.
- Add experimental API for reading and writing “mif3”-branded AVIF files behind the compilation flag AVIF_ENABLE_EXPERIMENTAL_METAV1.
- Implement avifImageScale() fallback when libyuv is not available.
- Partial import of libyuv to third_party/libyuv (new LICENSE).
- Add avifenc flag suffixes “:update” and “:u”. Quality-relative, tiling-relative and codec-specific flags can now be positional, relative to input files.
- Add experimental support for layered AVIF encoding in avifenc. Use the --layered flag to enable layered AVIF encoding. Layered AVIF has multiple layers, which works like frame of animated AVIF, and layers can be rendered in progressive manner on supported viewers (e.g. Chrome 94 or newer). Only aom supports layered AVIF encoding at the time of writing. Add --scaling-mode flag to set scaling mode of each layer. This part of AV1 encoder is not as thoroughly tested, so there are higher possibility encoder may crash when given certain configuration or input.
- Add imageSequenceTrackPresent flag to the avifDecoder struct.
- avifImageScale() function was made part of the public ABI.
- Add avif_cxx.h as a C++ header with basic functionality.
- Add enum aliases AVIF_COLOR_PRIMARIES_SRGB, AVIF_COLOR_PRIMARIES_BT2100, AVIF_COLOR_PRIMARIES_DCI_P3, AVIF_TRANSFER_CHARACTERISTICS_PQ.
- Add avifResult enum entry AVIF_RESULT_INTERNAL_ERROR.
- Require libyuv by default (but it can still be disabled with -DAVIF_LIBYUV=OFF).
- Add avifdec --icc flag to override the output color profile.
- Add experimental API for reading and writing 16-bit AVIF files behind the compilation flag AVIF_ENABLE_EXPERIMENTAL_SAMPLE_TRANSFORM.
- Add AVIF_CHROMA_SAMPLE_POSITION_RESERVED to avifChromaSamplePosition enum.
Changed since 1.0.0
- Update aom.cmd: v3.9.1
- Update avm.cmd: research-v7.0.1
- Update dav1d.cmd: 1.4.3
- Update libgav1.cmd: v0.19.0
- Update libjpeg.cmd: v3.0.3
- Update libxml2.cmd: v2.12.7
- Update libyuv.cmd: a6a2ec65
- Update mp4box.sh: v2.4.0
- Update rav1e.cmd: v0.7.1
- Update svt.cmd/svt.sh: v2.1.1
- Update zlibpng.cmd: zlib 1.3.1 and libpng 1.6.40
- AVIF sequences encoded by libavif will now also have the “avio” brand when there is at least one track made only of AV1 keyframes.
- Fix SVT-AV1 codec interface which was not setting video range at encoding.
- Any item ID being 0 in an “iref” box with version 0 or 1 is now treated as an error instead of being ignored.
- API calls now return AVIF_RESULT_OUT_OF_MEMORY instead of aborting on memory allocation failure.
- avifdec and avifenc: Change the default value of the --jobs option from 1 to “all”.
- Update avifCropRectConvertCleanApertureBox() to the revised requirements in ISO/IEC 23000-22:2019/Amd. 2:2021 Section 7.3.6.7.
- AVIF files with an exif_tiff_header_offset pointing at another byte than the first II or MM tag in the Exif metadata payload will now fail to be decoded. Set decoder->ignoreExif to true to skip the issue and decode the image.
- Fix memory errors reported in crbug.com/1501766, crbug.com/1501770, and crbug.com/1504792 by Fudan University.
- For codecs, AVIF_CODEC_* and AVIF_LOCAL_* are now merged into AVIF_CODEC_* that can only take the values: OFF, LOCAL or SYSTEM.
- For the gtest, jpeg, libsharpyuv, libxml2, libyuv and zlibpng dependencies, AVIF_LOCAL_* is now replaced by flags AVIF_* that can take the values: OFF, LOCAL or SYSTEM.
- src/reformat.c: Allocate the threadData array directly.
- AVIF_ENABLE_WERROR is set to OFF by default.
- Fix wrong alpha plane deallocation when decoded tile pixel format does not match reconstructed output image pixel format (b/320234262).
- Fix identical chunk skipping optimization when writing animation data (b/321189607).
- Fix ID selection for artificial grid alpha item when decoding a grid of tiles which each have an associated auxiliary alpha image item (https://crbug.com/oss-fuzz/65657).
- ext/libjpeg.cmd now pulls libjpeg-turbo instead of libjpeg and AVIF_JPEG=LOCAL now expects the library dependency in ext/libjpeg-turbo/build.libavif.
- Fix ‘iloc’ box parsing bugs that may have wrongly accepted, rejected or parsed some files with rare values of offset_size, length_size, base_offset_size and index_size.
- ‘infe’ boxes with an item_type different from ‘mime’ and without a null-terminated item_name are now considered invalid as per ISO/IEC 14496-12.
1.0.4 - 2024-02-08
Changed
- AVIF_ENABLE_WERROR is set to OFF by default.
- Fix wrong alpha plane deallocation when decoded tile pixel format does not match reconstructed output image pixel format (b/320234262).
- Fix identical chunk skipping optimization when writing animation data (b/321189607).
- Fix ID selection for artificial grid alpha item when decoding a grid of tiles which each have an associated auxiliary alpha image item (https://crbug.com/oss-fuzz/65657).
1.0.3 - 2023-12-03
Changed
- Rewrite the fix for memory errors reported in crbug.com/1501770.
- Fix memory errors reported in crbug.com/1504792 by Fudan University.
- src/reformat.c: Allocate the threadData array directly.
1.0.2 - 2023-11-16
Changed
- Update avifCropRectConvertCleanApertureBox() to the revised requirements in ISO/IEC 23000-22:2019/Amd. 2:2021 Section 7.3.6.7.
- Fix memory errors reported in crbug.com/1501766 and crbug.com/1501770 by Fudan University.
1.0.1 - 2023-08-29
Changed
- gdk-pixbuf: Explicitly pass link directories
- gdk-pixbuf: Fix build failure after imir.mode -> imir.axis rename
1.0.0 - 2023-08-24
With the 1.0.0 release, the ABI will be more stable from now on. Please note the allocation and initialization requirements for avifImage, avifDecoder, avifEncoder, and avifRGBImage in the “avif/avif.h” header.
List of incompatible ABI changes in this release:
- The clli member was added to the avifImage struct.
- The repetitionCount member was added to the avifEncoder and avifDecoder structs.
- The quality and qualityAlpha members were added to the avifEncoder struct.
- Check that functions returning pointers do not return NULL before accessing those pointers.
- Check the return value of avifEncoderSetCodecSpecificOption().
- The maxThreads member was added to the avifRGBImage struct.
- Check the return value of avifRGBImageAllocatePixels(), avifRWDataRealloc(), avifRWDataSet(), avifImageSetProfileICC(), avifImageSetMetadataExif() and avifImageSetMetadataXMP().
- The meaning of the keyframeInterval member of avifEncoder struct has changed slightly. When set to a value of “n”,
- Before: It forces a keyframe on every nth frame.
- After: Any set of “n” consecutive frame will have at least one keyframe (every nth frame may or may not be a keyframe).
Added
- Add STATIC library target avif_internal to allow tests to access functions from internal.h when BUILD_SHARED_LIBS is ON.
- Add clli metadata read and write support
- Add repetitionCount member to avifEncoder and avifDecoder structs to specify the number of repetitions for animated image sequences.
- Add quality and qualityAlpha to avifEncoder. Note: minQuantizer, maxQuantizer, minQuantizerAlpha, and maxQuantizerAlpha are deprecated. Code should be updated to set quality (and qualityAlpha if applicable) and leave minQuantizer, maxQuantizer, minQuantizerAlpha, and maxQuantizerAlpha initialized to the default values.
- The --target-size flag in avifenc was added to adapt the quality so that the output file size is as close to the given number of bytes as possible.
- Add the public API function avifImageIsOpaque() in avif.h.
- Add the public API functions avifImagePlane(), avifImagePlaneRowBytes(), avifImagePlaneWidth(), and avifImagePlaneHeight() in avif.h.
- Add experimental API for progressive AVIF encoding.
- Add API for multi-threaded YUV to RGB color conversion.
- Add experimental support for AV2 behind the compilation flag AVIF_CODEC_AVM. AVIF_CODEC_CHOICE_AVM is now part of avifCodecChoice.
- Add experimental YCgCo-R support behind the compilation flag AVIF_ENABLE_EXPERIMENTAL_YCGCO_R.
- Allow lossless 4:0:0 on grayscale input.
- Add avifenc --no-overwrite flag to avoid overwriting output file.
- Add avifenc --clli flag to set clli.
- Add support for all transfer functions when using libsharpyuv.
Changed
- Enable the libaom AV1E_SET_SKIP_POSTPROC_FILTERING codec control by default.
- Use the constant rate factor (CRF) instead of the constant quantization parameter (CQP) rate control mode with the SVT-AV1 encoder.
- Exif and XMP metadata is exported to PNG and JPEG files by default, except XMP payloads larger than 65502 bytes in JPEG.
- The --grid flag in avifenc can be used for images that are not evenly divided into cells.
- Apps must be built with libpng version 1.6.32 or above.
- Change the encoder to write the boxes within the “stbl” box in the order of stsd, stts, stsc, stsz, stco, stss.
- avifImageCopy() no longer accepts source U and V channels to be NULL for non-4:0:0 input if Y is not NULL and if AVIF_PLANES_YUV is specified.
- The default values of the maxQuantizer and maxQuantizerAlpha members of avifEncoder changed from AVIF_QUANTIZER_LOSSLESS (0) to AVIF_QUANTIZER_WORST_QUALITY (63). The behavior changed if minQuantizer and maxQuantizer are left initialized to the default values. Code should be updated to set the quality member. Similarly for the alpha quantizers and qualityAlpha.
- avifImageRGBToYUV() and avifImageYUVToRGB() handle avifImage bit depths 8, 10, 12 and now also 16. Files read by apps/shared/ can output 16-bit avifImage instances.
- Update aom.cmd: v3.6.1
- Update dav1d.cmd: 1.2.1
- Update libsharpyuv: 0.4.0
- Update rav1e.cmd: v0.6.6
- Update svt.cmd/svt.sh: v1.6.0
- Update zlibpng.cmd: zlib 1.2.13 and libpng 1.6.39
- avifImageCreate(), avifImageCreateEmpty(), avifEncoderCreate() and other internal functions now return NULL if a memory allocation failed.
- avifEncoderSetCodecSpecificOption() now returns avifResult instead of void to report memory allocation failures.
- At decoding, avifIOStats now returns the same values as at encoding.
- avifRGBImageAllocatePixels(), avifRWDataRealloc(), avifRWDataSet(), avifImageSetProfileICC(), avifImageSetMetadataExif() and avifImageSetMetadataXMP() now return avifResult instead of void to report memory allocation failures.
- avifReadImage(), avifJPEGRead() and avifPNGRead() now remove the trailing zero byte from read XMP chunks, if any. See avifImageFixXMP().
- Force keyframe for alpha if color is a keyframe.
- Write primaries and transfer characteritics info in decoded PNG.
- Add support for reading PNG gAMA, cHRM and sRGB chunks.
- The ‘mode’ member of the avifImageMirror struct was renamed ‘axis’.
- Change the type of the ‘depth’ parameter from int to uint32_t in avifFullToLimitedY(), avifFullToLimitedUV(), avifLimitedToFullY(), and avifLimitedToFullUV().
0.11.1 - 2022-10-19
Changed
- avifincrtest_helpers: Cast 64-bit offset to size_t
- avifmetadatatest: don't include avif/internal.h
- avifrgbtoyuvtest: skip if no libsharpyuv
- Disable tests that may fail if the codec is not aom (#1176)
0.11.0 - 2022-10-12
There are incompatible ABI changes in this release. The alphaRange member was removed from the avifImage struct. The chromaDownsampling and avoidLibYUV members were added to the avifRGBImage struct. The imageDimensionLimit member was added to the avifDecoder struct. avifImageCopy() and avifImageAllocatePlanes() signatures changed. It is necessary to recompile your code. Also check the return values of avifImageCopy() and avifImageAllocatePlanes().
Added
- Add man pages for avifenc and avifdec
- Add the avifChannelIndex type alias for enum avifChannelIndex
- Add avifChromaDownsampling enum
- Add chromaDownsampling field to avifRGBImage struct
- Add support for AVIF_RGB_FORMAT_RGB_565
- Add imageDimensionLimit field to avifDecoder struct
- Add autoTiling field to avifEncoder struct
- Add new avifResult codes AVIF_RESULT_CANNOT_CHANGE_SETTING and AVIF_RESULT_INCOMPATIBLE_IMAGE
- Add new enum constants AVIF_PIXEL_FORMAT_COUNT and AVIF_RGB_FORMAT_COUNT
- avifdec: Add --dimension-limit, which specifies the image dimension limit (width or height) that should be tolerated
- avifenc: Add --sharpyuv, which enables “sharp” RGB to YUV420 conversion, which reduces artifacts caused by 420 chroma downsampling. Needs libsharpyuv (part of the libwebp repository) at compile time.
- avifenc: Add --ignore-exif and --ignore-xmp flags.
- avifenc: Add --autotiling, which sets --tilerowslog2 and --tilecolslog2 automatically.
- avifenc: Input Exif orientation is converted to irot/imir by default.
Changed
- Fix memory leaks of metadata on avifenc exit
- Update the handling of ‘lsel’ and progressive decoding to AVIF spec v1.1.0
- Treat an absent lsel and layer_id == 0xFFFF equivalently for backward compatibility with earlier drafts of AVIF spec v1.1.0
- Set libavif's own default value of cfg.rc_end_usage for libaom
- Fix https://github.com/AOMediaCodec/libavif/issues/953
- Set the libaom-specific option -a tune=ssim by default
- Bump cmake_minimum_required from 3.5 to 3.13
- Fix https://crbug.com/oss-fuzz/48135
- Use several new libyuv functions in reformat_libyuv.c
- Fix SVT-AV1's issue 1957 related to uninitialized variables crashing the encoder
- Fix https://github.com/AOMediaCodec/libavif/issues/787
- Update aom.cmd: v3.5.0
- Update rav1e.cmd: v0.5.1
- Update svt.cmd/svt.sh: v1.2.1
- Update libgav1.cmd: v0.18.0
- Update libyuv.cmd: f9fda6e7 (version 1844)
- avifImageCopy() and avifImageAllocatePlanes() now return avifResult instead of void to report invalid parameters or memory allocation failures.
- avifImageRGBToYUV() now uses libyuv fast paths by default. It may slightly change conversion results. The old behavior can be restored by setting avifRGBImage::chromaDownsampling to AVIF_CHROMA_DOWNSAMPLING_BEST_QUALITY and avifRGBImage::avoidLibYUV to AVIF_TRUE.
- avifRGBImage::chromaUpsampling now only applies to conversions that need upsampling chroma from 4:2:0 or 4:2:2 and has no impact on the use of libyuv. Set avifRGBImage::avoidLibYUV accordingly to control the use of libyuv.
- avifenc: Set the YUV format to 4:0:0 for grayscale PNGs
- Support updating encoder settings and codec-specific options during encoding
- Disable AVIF_STRICT_CLAP_VALID and AVIF_STRICT_PIXI_REQUIRED in the JNI wrapper
- avifdec: Return proper exit code in “info” mode
- In avifenc and avifdec, treat all arguments that start with ‘-’ as options
- Fix https://github.com/AOMediaCodec/libavif/issues/1086
- Exif and XMP metadata is imported from PNG and JPEG files.
- avifImageSetMetadataExif() parses the Exif metadata and converts any Exif orientation found into transformFlags, irot and imir values.
- Write ‘auxi’ box for animated images with alpha channel
- Write ‘auxv’ as handler_type for alpha channel track
- Use PNG_COLOR_TYPE_GRAY for 8-bit grayscale output
- Replace repeated subtraction by modulo in calcGCD (fix b/246649620)
- Change avifImageCreate to take uint32_t instead of int parameters
- When writing an image sequence, check if it's safe to cast width and height to uint16_t
- Allow clamped grid cells in avifEncoderAddImageGrid()
Removed
- alphaRange field was removed from the avifImage struct. It it presumed that alpha plane is always full range.
- The avifCodecConfigurationBox struct becomes a private type for libavif internal use
0.10.1 - 2022-04-11
Changed
- tests/docker/build.sh: Build SVT-AV1 using cmake and ninja directly
- Fix a Visual Studio 2017 compiler warning in src\reformat.c: warning C4204: nonstandard extension used: non-constant aggregate initializer
- Fix the help message of avifdec: --index takes a value
0.10.0 - 2022-04-06
There is an incompatible ABI change in this release. New members were added to the avifDecoder and avifRGBImage structs. It is necessary to recompile your code.
Added
- Support F16 Half Float conversion in avifRGBImage: new isFloat member
- Incremental decoding of AVIF grid tiles: new allowIncremental member in avifDecoder and new avifDecoderDecodedRowCount() function
- Support parsing of version 3 of ItemInfoEntry
- Add new avifResult code AVIF_RESULT_OUT_OF_MEMORY
- Document the “[Strict]” prefix in error strings
- Document that SVT-AV1 doesn't support lossless yet
- CI: Add CIFuzz integration
- Add Docker build CI pipeline
- Add SVT-AV1 to CI and build scripts
- ci.yml: Build examples and apps
Changed
- Print the item type in the diagnostic messages for missing mandatory av1C or pixi property
- Update aom.cmd: v3.3.0
- Update dav1d.cmd: 1.0.0
- Update libgav1.cmd: 0.17.0
- Update rav1e.cmd: 0.5.0
- Update svt.cmd/svt.sh: v0.9.1
- Update zlibpng.cmd: zlib v1.2.12
- findrav1e: add LDFLAGS to LIBRARIES
- rav1e: add bcrypt.lib to list of extra libs
- Fix y4m read/write for images of non-standard dimensions
- Fix y4mRead() and y4mWrite() for 4:0:0
- Fix compilation with Clang 13 and 14
- Remove the obsolete script fuzz.sh
- Support local android builds for libgav1
- Add Android JNI bindings
- Delay failures of AV1 codecs not existing to frame decoding, to allow libavif to perform AVIF parsing without any AV1 codecs
- Change encoder speed in gdk-pixbuf plug-in
- Fix compilation with 1755 <= LIBYUV_VERSION < 1774
- Remove JSON-based tests (as they are unreliable), along with associated helper code (cJSON, compare)
- CMakeLists.txt: Move codec enabled message after check passed
- Fix alpha copy in aomCodecEncodeImage()
- Support SVT-AV1 v0.9.0 or later
- Call svt_av1_get_version() for SVT-AV1 v0.9.0 or later
- Handle avifArrayCreate() failures
- Only consider a frame index to be a keyframe if all tiles/planes are sync frames
- Move checks to avifAreGridDimensionsValid()
- avifArrayPop() should zero the popped element
- avifDecoderReset() should not return AVIF_FALSE
- Handle avifDecoderDataCreateTile() failures
- Fix endian dependent parameters to avifRWStreamWrite
- Mark the input images of an image grid as hidden
- Write ccst box in Sample Entry for animated images
- Add iso8 to compatible_brands for animated images
- Compare with snapshot of AOM_EXT_PART_ABI_VERSION
- Handle the new AOM_IMG_FMT_NV12 enum conditionally in a switch statement in aomCodecGetNextImage()
- Fix avifpng.c for libpng 1.4
- Fix -Wformat / -Wformat-non-iso on MinGW UCRT
- Replace some memcpy calls with struct assignments
- Remove unnecessary memcpy() calls in src/utils.c
- Split CMakeLists.txt into tests/CMakeLists.txt
- Use bilinear chroma upsampling in libyuv when possible
- Call libyuv functions to convert 10bpc YUV to 8bpc RGB
- Prepare avif example for non-aborting avifAlloc()
- Handle the tileRowsLog2 and tileColsLog2 members of avifEncoder correctly for SVT-AV1.
0.9.3 - 2021-10-20
Added
- Support for progressive AVIFs and operating point selection
- Add automatic tile scaling to the item‘s ispe or track’s dims
- Add diagnostic messages for AV1 decode failures
- avifdec: Add PNG compression level arg
- Make image size limit configurable, expose to avifdec
- Add the AVIF_STRICT_ALPHA_ISPE_REQUIRED flag
Changed
- Mandate ispe and disallow zero width or height (#640).
- Re-map libavif speed 7-10 to libaom speed 7-9 (#682)
- Refer to https://aomedia-review.googlesource.com/c/aom/+/140624
- If you were using libaom with the following avif speed setting:
- speed 0-6: no change is needed
- speed 7: change to speed 6 for the same results
- speed 8-9: re-test and re-adjust speed according to your app needs
- Update aom.cmd: v3.2.0
- Update dav1d.cmd: 0.9.2
- Update svt-av1.cmd: v0.9.0
- Pass TestCase's minQuantizer, maxQuantizer, speed to encoder.
- Regenerate tests.json
- Disable JSON-based tests for now, the metrics are inconsistent/unreliable
- Set diagnostic message for aom_codec_set_option()
- Re-map libavif-libaom speed settings (#682)
- Bump of version in CMakeLists.txt was forgotten
- avifdec: Better message for unsupported file extension
- Do not copy input image when encoding with libaom unless width or height is 1
- Fix the comment for AVIF_STRICT_PIXI_REQUIRED
- Update libavif.pc.cmake (#692)
- In 32-bit builds set dav1d's frame_size_limit setting to 8192*8192
- Allocate alpha alongside YUV (if necessary) during y4m decode to avoid incorrect alphaRowBytes math
- Change avif_decode_fuzzer to be more like Chrome
- Update codec_dav1d.c for the new threading model
- Generalized ipco property deduplication
- Rename avifParseMoovBox to avifParseMovieBox for consistency
- Simplify idat storage for avifMeta structure (#756)
- Fix oss-fuzz coverage build failure of dav1d
- Redesign AVIF_DECODER_SOURCE_AUTO to honor the FileTypeBox's major brand
- Use “C420” as default Y4M color space parameter
0.9.2 - 2021-06-23
Added
- avifenc, avifdec: Allow “-j all” to automatically use all of the cores on the machine (#670)
Changed
- Refactor imir implementation to match HEIF Draft Amendment 2 (#665)
- Merge avifCodec's open call with its getNextImage call to avoid codec init during parse, and simplify the codec API (#637)
- Update aom.cmd: v3.1.1 (#674)
- Update svt-av1: v0.8.7 (#627)
- Make tests/compare.h and tests/testcase.h C++ safe (#678)
- Print width and height as %ux%u instead of %u/%u (#676)
- Allocate codec->internal->svt_config statically (#675)
- Cleanup related to avifDiagnosticsClearError() (#673)
- Cleanup avifutil.h comment to match libavif style (#671)
- Fix the clang -Wunused-macros warning (#672)
- Check for int32_t overflows in ‘clap’ code (#663)
- Have avifdec print chroma sample position for 420 (#666)
- Enable CMake configs in VCPKG mode (#659)
- Avoid multiplying widthN and heightN by 2 (#662)
- Correct AVIF_PIXEL_FORMAT_NONE handling logic (#654)
- Cast extent->offset (a uint64_t) to size_t safely (#660)
- Disallow negative clap width or height (#656)
- Check for int32_t cast and unsigned add overflows (#655)
- Some straightforward changes to clapFraction code (#653)
- Fix box name of avifParseChunkOffsetBox (#652)
- No need to pass diag to functions that have ‘data’ (#651)
- Simplify the assertion in avifROStreamStart() (#650)
- Don't clear error in avifEncoderSetCodecSpecificOp (#648)
- Simplify avifCodecConfigurationBoxGetFormat (#646)
- Print the fraction in “not an integer” messages (#641)
- Fix a typo in the diagnostic context for ‘ipco’ (#644)
- Remove const from non-pointer function parameters (#634)
- Declare the param of avifDumpDiagnostics as const (#633)
- Adjust gdk-pixbuf loader for new API change (#668)
- Fix gdk-pixbuf loader install path (#615)
0.9.1 - 2021-05-19
Added
- Added strict mode/flags (enabled by default):
AVIF_STRICT_PIXI_REQUIRED
, AVIF_STRICT_CLAP_VALID
- avifdec: Added
--no-strict
to disable all strict flags - avifdec: Added
-r
(--raw-color
), which avoids multiplying against AVIF alpha channel before packing into non-alpha formats (JPEG) - avifenc: Recognize the Y4M format string “C420mpeg2”
- avifenc: Add
--crop
convenient alternative arg to the difficult-to-use --clap
arg - avifenc: New default for
--yuv
: "auto"
, which will use a source JPEG's internal YUV format instead of YUV444, if detected- Uses: Prevent colorspace conversion when reading from JPEG if possible (tongyuantongyu)
- avifenc/avifdec: Add helpful values/calculations when dumping clap box
- Added avifDiagnostics, which allows for a detailed, freeform error string upon decode or encode error
- Create helper avifCropRect struct and methods for helping to manipulate/populate/validate avifCleanApertureBox
- Added ability to set codec-specific options for color or alpha only
- Support for libaom's ALL_INTRA mode (if available)
- Create avifDecoder.imageCountLimit as a sanity check against malformed files
- SVT: Image sequence encoding support (tongyuantongyu)
- Added rav1e to AppVeyor builds
Changed
- avifenc/avifdec: Link AOM_LIBRARIES and use CXX if vmaf is present (1480c1)
- Ensure that an AVIF has a ftyp box, and based on ftyp, has other appropriate toplevel boxes present as well
- Avoid linking against libyuv if it is too old / incompatible
- Always require a primary item when decoding items
- Add some strictness around ipma box parsing (version/flags tuples must be unique across ipma boxes in a file)
- Fix alpha grids by properly writing alpha grid metadata payload
- A HandlerBox (hdlr) of type ‘pict’ must be the first box within the MetaBox (meta)
- Add some typedefs for various flag decls in avif.h to self-document which flags should be used in which function arguments
- When encoding single-frame images using libaom, clean up the encoder immediately after encoding the frame to cut down on resources high watermarks
- Fail on reformat Identity (MC=0) with subsampling (not using YUV444)
- Warn if alpha is limited range (deprecated)
- Validate the first_chunk fields in the stsc box
- In libaom all intra mode, set cq-level for user
- Check the return values of some aom_codec_ calls and add diagnostics output (wantehchang)
- Use aom_codec_set_option() if available (allows for future compat with libaom 3.0+ advanced features)
- rav1e: Use cargo cinstall in local builds to ensure consistency in target output, as cbuild no longer builds directly into target/release
- Tweaks to compiler flags (analyze related)
- Use libyuv BT.709 & 2020 full range YuvConstants (wantehchang)
- Multiply color with alpha for opaque RGB format during conversion (see #520)
- Switch docker to ubuntu 20.04, fix tzdata install (paskal)
- Added an “Understanding maxThreads” explanatory comment block in avif.h
- Minor fixes to support AVIF_CODEC_AOM_ENCODE
- Various minor code/comments cleanup
- CI tweaks, macOS build, and caching / speed increases (EwoutH)
- Update aom.cmd: v3.1.0
- Update dav1d.cmd: 0.9.0
- Update libgav1: v0.16.3
- Update libyuv.cmd: 2f0cbb9
0.9.0 - 2021-02-22
Added
- Image grid encoding
- Premultiplied alpha support (tongyuantongyu)
- avifenc: Image grid encoding (
-g
, --grid
) - avifenc: Harvest framerate from y4m headers as the “default”, if present
- avifenc: Recognize the Y4M format string “C420mpeg2” (wantehchang)
- Basic deduplication when writing mdat chunks
Changed
- avifenc: Adjusted min/max/speed/fps defaults
- Better handling for export headers (tongyuantongyu)
- Use procedure specified in H.273 to quantize YUV (tongyuantongyu)
- Impose a maximum of 4096 bytes on searchSampleSize (wantehchang, fixes oss-fuzz perf issue / timeout)
- Update aom.cmd: v2.0.2
- Update dav1d.cmd: 0.8.2
- Update libgav1.cmd: 4a89dc3 / lts_2020_09_23
- Update rav1e.cmd: 0.4
- Update svt.cmd/svt.sh: v0.8.6
- Force libjpeg to output in RGB Colorspace (bugfix)
- Minor other compilation/linking/formatting/comment fixes
0.8.4 - 2020-11-23
Added
- YCgCo support (full-range only, wantehchang)
- Expose
maxThreads
to avifDecoder
, add --jobs
to avifdec
- Add
avifDecoderNthImageMaxExtent()
streaming helper function to determine future reads - Create AVIF_VERSION_DEVEL, which allows for in-development changes to be detected during the build
- New avifResult values:
AVIF_RESULT_INVALID_ARGUMENT
, AVIF_RESULT_NOT_IMPLEMENTED
Changed
- Read/write one of each type of colr box, as allowed in HEIF 6.5.5.1 Amendment 3 (also now supported in libheif)
- avifenc/avifdec: Now guesses input file format by header inspection instead of extension
- avifenc: Fix y4m/stdin corner cases when trying to detect a “single image” encode
- Add some protections against improperly using AVIF_ADD_IMAGE_FLAG_SINGLE
- imir transformation: Disambiguate all usages of “vertical” and “horizontal” across comments and tooling
- Print MC value when warning resetting to defaults (wantehchang)
- Fix grid image rows/cols parsing on invalid data
- Allow override of HAVE_POISON_SYSTEM_DIRECTORIES_WARNING (bnason-nf)
- Lots of comments clarifications in avif.h
0.8.3 - 2020-11-09
Added
- SVT-AV1 encode support (jonsneyers)
- Basic libyuv support (8bpc only, see usage/limitations in avif.h)
- Refactor avifChromaUpsampling for ease in high level decision making and adding more filters in the future (minor, benign breaking change due to enum reordering)
- New CMake options:
AVIF_CODEC_AOM_DECODE
, AVIF_CODEC_AOM_ENCODE
- New examples in
examples/
dir (encode, decode_file, decode_memory, decode_streaming) to replace stale, not-compiled examples in README
Changed
- avifenc: Explicitly signal SRGB CP/TC if the source image and user don't specify, and no ICC profile is present
- Set g_lag_in_frames to 1 if encoding single image (encode memory optimization, wantehchang)
- Early-out with OK in avifParse() if the expected top-level boxes were already parsed (streaming optimization)
- Check increading item_ID and ipmaSeen in ipma box (wantehchang)
- Fail gracefully in avifEncoderAddImage() if libavif wasn't compiled with encoding support
- Add size_t casts to fix MSVC x86 compiler warnings (wantehchang)
- Link with {ZLIB_LIBRARY} after ${PNG_LIBRARY} (wantehchang)
- Fix a crash in avifJPEGRead() on fopen() failure (wantehchang)
- No need to include
${ZLIB_INCLUDE_DIR}
(wantehchang) - Put the value of
ZLIB_INCLUDE_DIR
in the cache (wantehchang) - Don't set image->alphaRange to yuvRange (wantehchang)
- Expose chroma sample position in decoded images (wantehchang)
- avifDecoderNthImage: tighten decoder flush (wantehchang)
- Cleanup avifIOReadFunc comments for clarity (wantehchang)
- Minor code cleanup
- Minor CMake cleanup
Removed
- Remove disableGridImages from avifDecoder (wantehchang)
0.8.2 - 2020-10-14
Added
avifIO
reader API: allowing for parsing / image decoding without having the entire AVIF payload yet- Codec-specific options plumbing for advanced encoding settings
- Add libaom codec-specific options (wantehchang)
- avifenc: Allow endusers to ignore an AVIF's ICC profile during conversion (
--ignore-icc
) - avifenc: Allow the setting/overriding of XMP, Exif, and ICC profiles
- Add the
disableGridImages
setting to avifDecoder
- Add AVIF_FMT_ZU to fix compiling with non-standard/old compilers
- Add
AVIF_ENABLE_WERROR
(ON by default) - Add
AVIF_ENABLE_COVERAGE
for basic llvm coverage report generation
Changed
- Support multiple extents in an ItemLocationBox
- Store all alpha payloads before color payloads in mdat
- Perform 0.5 UV bias with integers, as 128/512/2048 aren't exactly 0.5, but are expected to behave as such
- Avoid libpng's complaints about specific ICC profiles
- Disable receiving one-frame-per-layer when decoding scalable AVIFs with aom and dav1d
- Fix incorrect 8-to-16 monochrome YUV conversion
- Set max image size to 16384 * 16384
- Remove range and sample position from avifImageStealPlanes()
- Ensure only one of each mandatory-unique box in a meta box exists
- Ensure each item ID is cited once in an iloc box
- Sanity check merged extents item size against the file size
- Various image grid bugfixes
- Error out with AVIF_RESULT_REFORMAT_FAILED if request uses an unsupported MC
- Fix memory leak in avifenc when encoding image sequences
- Move oss-fuzz fuzzer implementation into
tests/oss-fuzz
- avifdec: actually propagate the commandline codecChoice to the avifDecoder
- Fix an infinite loop in codec_dav1d on a carefully-malformed AV1 payload
- Fix a few issues with warnings in gcc/clang
- Various comments tweaks
0.8.1 - 2020-08-05
Added
- Add
ignoreAlpha
field to avifRGBImage (linkmauve) - Save support in gdk-pixbuf component (novomesk)
Changed
- Only ever create one iref box, filled with multiple cdsc boxes (#247)
- Fix incorrect 16-to-8 monochrome YUV conversion
- Make decoding optional in CMake, like encoding is
- Include avif INTERFACE_INCLUDE_DIRECTORIES first (cryptomilk)
- Set C standard to C99, adjust flags for dav1d (1480c1)
- Minor cleanup/fixes in reformat.c (wantehchang)
- Fix a crash in the gdk-pixbuf loader, removed unnecessary asserts (novomesk)
0.8.0 - 2020-07-14
Added
- Monochrome (YUV400) support **
- All encoding/decoding and internal memory savings are done/functional
- libaom has a bug in chroma_check() which crashes when encoding monochrome, to be fixed in a future (>v2.0.0) version
- rav1e didn't implement CS400 until rav1e v0.4.0
- libavif safely falls back to YUV420 when these earlier codec versions are detected
- NOTE: If you want to do heavy monochrome testing, wait for newer versions to libaom/rav1e!
- Image sequence encoding support
- Required medium-sized refactors in the codec layers
- Image sequences (tracks) now fully support all metadata properly (Exif/XMP/transforms)
- avifenc can now encode a series of same-sized images with a consistent framerate, or each with their own custom duration
- Bilinear upsampling support
- avifenc: Add --ignore-icc, which avoids embedding the ICC profile found in the source image
- avifdec: Add --info, which attempts to decode all frames and display their basic info (merge of avifdump)
- avifenc: add --tilerowslog2 and --tilecolslog2 (wantehchang)
- Added
contrib
dir for any unofficially supported code contributions (e.g. gdk-pixbuf)
Changed
- CICP Refactor (breaking change!)
- Remove most references to “NCLX”, as it is mostly an implementation detail, and the values are really from MPEG-CICP
- Eliminate avifProfileFormat: having an ICC profile is not mutually exclusive with signaling CICP
- CICP is now always available in an avifImage, set to unspecified by default
- Added --cicp as an alias for --nclx (semi-deprecated)
- Setting CICP via avifenc no longer overrides ICC profiles, they co-exist
- Simplified avifenc argument parsing / warnings logic
- avifenc/avifdec/avifdump now all display CICP when dumping AVIF information
- nclx colr box contents are guaranteed to override AV1 bitstream CICP (as MIAF standard specifies)
- Added comments explaining various decisions and citing standards
- Removed ICC inspection code regarding chroma-derived mtxCoeffs; this was overdesigned. Now just honor the assoc. colorPrimaries enum
- Reworked all examples in the README to reflect the new state of things, and clean out some cruft
- Harvest CICP from AV1 bitstream as a fallback in avifDecoderParse() if nclx box is absent
- All data other than actual pixel data should be available and valid after a call to avifDecoderParse()
- Refactor avifDecoder internal structures to properly handle meta boxes in trak boxes (see avifMeta)
- Update libaom.cmd to point at the v2.0.0 tag
- Update dav1d.cmd to point at the 0.7.1 tag
- Re-enable cpu-used=7+ in codec_aom when libaom major version > 1
- Memory allocation failures now cause libavif to abort the process (rather than undefined behavior)
- Fix to maintain alpha range when decoding an image grid with alpha
- Improvements to avifyuv to show drift when yuv and rgb depths differ
- Remove any references to (incorrect) “av01” brand (wantehchang)
- Set up libaom to use reduced_still_picture_header (wantehchang)
- Use libaom cpu_used 6 in “good quality” usage mode (wantehchang)
- Update avifBitsReadUleb128 with latest dav1d code (wantehchang)
- Set encoder chroma sample position (wantehchang)
0.7.3 - 2020-05-04
Added
- avifenc: Lossless (--lossless, -l) mode, which sets new defaults and warns when anything would cause the AVIF to not be lossless
Changed
- Minor cleanup for -Wclobbered warnings
- Minor fixes to README and code (fallout from enum rework)
- Protect against oversized (out of bounds) samples in avif sample tables
- Optimization: avoid AV1 sample copying when feeding data to dav1d
0.7.2 - 2020-04-24
Added
- Recognize extensions with capital letters / capslock
- Proper support for AVIF_NCLX_MATRIX_COEFFICIENTS_IDENTITY
Changed
- Large nclx enum refactor (breaking change), reworking all 3 enums to better match AV1 codec enums
- Fixes to ‘essential’ item properties (marking av1C as essential, ignoring any items containing unsupported essential props)
- avifenc - Allow --nclx to override embedded ICC profiles (with a warning), instead of --nclx being ignored
- avifenc - Choose high-quality-but-lossy QP defaults, and a default speed of 8
- avifdump - Fix format specifiers for 32bit
- Now prioritizing libaom over rav1e when both are present
- Remove
-Wclobbered
dodging (volatile) and instead just disable the warning in avifpng/avifjpeg - avifyuv: extra testing modes
- Cleanup to avifCodecVersions()
- Reorganize iccjpeg code back into its own files for licensing conveniences
0.7.1 - 2020-04-16
Changed
- avifenc: Set nclx/range values in avifImage earlier so proper YUV coefficients are used when converting JPEG/PNG
0.7.0 - 2020-04-16
Added
- avifenc and avifdec JPEG support
- Docker test script to build avifenc + deps in a shared libs (distro-like) env
- Added simple
avifdump
tool for aiding in AVIF debugging - Added some comments in
avif.h
to clarify avifDecoderSetSource()
usage
Changed
- avifRange cleanup/refactor (breaking change)
- avifenc now has
-r
to set YUV range (when using JPEG/PNG), --nclx
now takes 3 arguments as a result
0.6.4 - 2020-04-14
Added
- Added
avifDecoderNthImageTiming()
for querying frame timing without needing to decode the frame - Added some comments explaining
avifDecoderSetSource()
Changed
- Fix clang warning (switch clamp to min)
- Fix a few clang analyzer issues
- Avoid incorrect YUV range cast
- Call dav1d_data_unref in dav1dCodecDestroyInternal (wantehchang)
- Declare some avifSampleTable * pointers as const (wantehchang)
- Update to cJSON v1.7.13 (wantehchang)
- Minor code cleanup
0.6.3 - 2020-03-30
Changed
- Avoid throwing away const unnecessarily in
avifROStreamReadString()
- Re-enable a bunch of clang warnings
- Set dav1dSettings.frame_size_limit to avoid OOM (wantehchang)
- Refactor write.c to use a similar Data/Item design as read.c
- YUV to RGB optimizations
0.6.2 - 2020-03-11
Changed
- Fix 16bpc PNG output
- Compile fixes to avoid -Wclobbered in PNG code
- GitHub automatic deployment from AppVeyor (EwoutH)
0.6.1 - 2020-03-11
Added
- PNG support for avifenc/avifdec
Changed
- Fixed Clang10 build warning
- Fix SOVERSION in cmake (cryptomilk)
- Minor tweaks to avifBool usage (wantehchang)
0.6.0 - 2020-03-09
Added
avifRGBImage
structure and associated routines (BREAKING CHANGE)- avifImage alphaRange support
- Support pasp, clap, irot, imir metadata for encode/decode
Changed
- Large RGB conversion refactor (BREAKING CHANGE), see README for new examples
- Minor fixes to make Clang 10 happy
- pkg-config fixes
- Lots of minor cleanup in code/CMake (wantehchang)
- Fix to NCLX color profile plumbing (ledyba-z)
- Cleanup unnecessary avifBool ternary expressions
- Update default dav1d version to 0.6.0
- Update default rav1e version to v0.3.1
0.5.7 - 2020-03-03
Added
- libgav1 decode codec support. (wantehchang @Google)
- Expose codec selection to avifdec/avifenc, speed to avifenc
- Image grid support (Summer_in_Tomsk_720p_5x4_grid)
minQuantizerAlpha
/maxQuantizerAlpha
support in avifEncoder, avifenc- 444alpha support in y4m layer (avifenc, avifdec)
- pkg-config support (cryptomilk)
- Proper support of NCLX matrix coefficients enum (link-u)
Changed
- AppVeyor builds now compile with dav1d (EwoutH)
- Lots of minor CMake/code cleanup (wantehchang @Google)
- cJSON license note for aviftest (wantehchang @Google)
0.5.6 - 2020-02-19
Added
- Added CMake Find modules for aom, dav1d, rav1e (cryptomilk)
Changed
- use right-most and bottom-most UV pixels in images with odd-dimensions (ledyba-z)
- avoid libaom crash when encoding >8bpc images at high speed
0.5.5 - 2020-02-13
Added
- Enable still picture mode with rav1e >= 0.3.0 (cryptomilk)
- Basic test suite (aviftest, rough draft)
Changed
- Explicitly cast unorms to float during YUV conversion, fixing clang warning
- Optimize SampleSizeBox parsing when sample_size>0, fixes OOM oss-fuzz issue #5192805347753984
- Fix memory leak when using avifDecoderReset(), fixes oss-fuzz issue #5770230506979328
- Update default rav1e version from 0.2.1 to 0.3.0
- Remove a null check for codec->internal->image (wantehchang)
0.5.4 - 2020-01-21
Changed
- Fix monochrome inputs on avifImageCopy. Monochrome still isn't really a first-class citizen in libavif, but this should at least honor the incoming data better.
- Updated README‘s Basic Decoding section reminding of avifDecoderRead’s tradeoffs
- build: avoid -ldl if not required or not supported (jbeich)
- apps: convert ADVANCE to an expression (jbeich)
0.5.3 - 2019-12-03
Added
- Honor CMake's builtin
CMAKE_SKIP_INSTALL_RULES
Changed
- avifenc - Removed accidental double-delete of avifImage when failing to read a y4m file input
- Round dimensions down when decoding subsampled YUV with odd dimensions
0.5.2 - 2019-11-23
Changed
- Fix incorrect free in 0-case for
avifRWDataSet()
0.5.1 - 2019-11-21
Changed
- Fix expectations for Exif payload to better match normal usage
0.5.0 - 2019-11-21
Added
- Define version and SO-version for shared library
- Use -DBUILD_SHARED_LIBS=OFF for building a static lib
- avifImage can now hold Exif and XMP metadata (
avifImageSetMetadataExif
, avifImageSetMetadataXMP
) - Support for reading/writing Exif and XMP items
- Now tracking idat boxes across meta boxes
- Support for iloc construction_method 1 (idat)
Changed
- Proper handling of the primary item box (pitm) on read
- avifROStreamReadString() now allows string skipping by passing a NULL output buffer
- Updated README to show Exif/XMP support
0.4.8 - 2019-11-19
Added
- avifEncoder now has a speed setting
- codec_aom only flushes encoder when necessary (avoids lost frame packets)
- shared library compilation (build shared by default, use
-DAVIF_BUILD_STATIC=1
for static lib) - make install support
- cmake fixes/support for find_package (cryptomilk)
Changed
- Updated libaom to more recent SHA in aom.cmd
- Tweaked AVIF_LOCAL_AOM settings to play nice with libaom‘s usage of CMake’s option()
- Remove all libaom special cases from libavif's CMakefiles, and have it work the same way dav1d and rav1e do
- Minor cleanup
0.4.7 - 2019-11-11
Changed
- Fix memory leak in rav1e codec (PR20, AurelC2G)
- Bump rav1e version in rav1e.cmd, implement
avifCodecVersionRav1e()
- Display versions in avifenc and avifdec
0.4.6 - 2019-10-30
Changed
- Fix rav1e build on Linux x64, and eliminate pseudo-dependency on cargo-c
0.4.5 - 2019-10-30
Changed
- Fix rav1e codec's alpha encoding (monochrome asserts, might be unsupported still)
0.4.4 - 2019-10-30
Changed
- Fix QP range for rav1e encodes (rav1e uses [0-255], not [0-63])
- Distribute out and share code populating av01 config box across codecs
0.4.3 - 2019-10-28
Added
- rav1e codec support (encode-only)
rav1e.cmd
and dav1d.cmd
to ext
Changed
- All codecs can coexist peacefully now, and can be queried for availability or specifically chosen at encode/decode time
- Updated README to indicate changes to CMake which facilitate codec reorg
0.4.2 - 2019-10-17
Changed
- Populate nclx box inside of OBU in addition to AVIF container
0.4.1 - 2019-10-17
Added
- Added
containerDepth
to avifDecoder for surfacing 10bpc/12bpc flags from av1C boxes, if present - Added
avifCodecVersions()
for getting version strings of internal AV1 codecs
Changed
- Fixed warning with CHECK macro (additional semicolon)
0.4.0 - 2019-10-02
Added
- exposed util functions:
avifFullToLimitedY
, avifFullToLimitedUV
, avifLimitedToFullY
, avifLimitedToFullUV
, avifPrepareReformatState
Changed
- Renamed ispeWidth/ispeHeight to containerWidth/containerHeight; they now can hold tkhd's width/height
- Split avifImageYUVToRGB into faster internal functions (estimated gain: 3.5x)
- Fixed a few memory leaks, one in the README, one in codec_dav1d (AurelC2G)
0.3.11 - 2019-09-26
Added
- Exposed ispeWidth/ispeHeight to decoder if decoding items with an associated ispe box
- Now parsing/tracking sample description formats to filter non-av01 type tracks
- Allow brand ‘av01’ to be decoded
Changed
- Fixed bug in sync sample table element sizing
- Pass through starting sample index to codec when flushing with NthImage
0.3.10 - 2019-09-26
Added
- stss box parsing for keyframe information
- avifBool avifDecoderIsKeyframe(avifDecoder * decoder, uint32_t frameIndex);
- uint32_t avifDecoderNearestKeyframe(avifDecoder * decoder, uint32_t frameIndex);
- avifResult avifDecoderNthImage(avifDecoder * decoder, uint32_t frameIndex);
- aviffuzz prints keyframe information as it repeatedly decodes
Changed
- internally renamed codec function “decode” to “open”, as that's all it does
- dav1d codec's open function no longer does an initial unnecessary feed
- avifCodecDecodeInput now stores an array of avifSample which know if they're keyframes
- moved codec flushing code into avifDecoderFlush() so it is available to avifDecoderNthImage
- ptsInTimescales is now calculated independently of frame decode order
0.3.9 - 2019-09-25
Changed
- Split avifRawData and avifStream into read-only (const) and read/write versions, updated code accordingly
- Fix a few clang/macOS warnings
0.3.8 - 2019-09-04
Changed
- Reverted codec_aom and libaom to use previous SHA (v1.0.0-errata1 is ancient)
0.3.7 - 2019-09-04 - DO NOT USE THIS VERSION
Added
- Check for proper width/height/depth when decoding alpha with dav1d, matching libaom's impl
Changed
- Updated codec_aom and libaom to use v1.0.0-errata1
0.3.6 - 2019-07-25
Added
- Exposed tile encoding to avifEncoder
0.3.5 - 2019-07-25
Changed
- Fixed copypasta bug in libaom encoding quantizer setup
0.3.4 - 2019-07-25
Added
- When the AVIF container does not contain a color profile, fallback to the color OBU's nclx
0.3.3 - 2019-07-24
Added
- new helper function
avifPeekCompatibleFileType()
- expose ioStats on avifDecoder again (currently only interesting when reading items)
Changed
- Fixed some warnings (removed unused variables and a bad cast)
- Add a define in dav1d layer for supporting older dav1d codecs
- Enabled tons of warnings, and warnings-as-errors; Fixed associated fallout
- codec_dav1d: disambiguate “needs more data” and “no more frames” in feed data pump
0.3.2 - 2019-07-23
Added
- Added
ext/aom.cmd
to perform a local checkout of the aom codebase, as an alternative to a real submodule. This allows downstream projects to use libavif without recursive submodule issues. - AppVeyor and Travis scripts now explicitly clone libaom into ext/ as an alternative to a submodule.
Changed
- Remove
ext/aom
as a submodule. If libavif users want to build aom from ext/, they must enable AVIF_BUILD_AOM
and supply their own local copy. - Move the handful of public domain gb_math functions used by colr.c and eliminate the dependence on the gb library
- Detect when libaom or libdav1d is being included by a parent CMake project and allow it
- Offer libavif's include dir alongside the library in CMake (target_include_directories)
0.3.1 - 2019-07-22
Changed
- Moved dependency on libm to avif executables, instead of directly on the library
- Minor changes to README examples
0.3.0 - 2019-07-22
Added
- new CMake option
AVIF_CODEC_AOM
to enable/disable the usage of AOM's codec (default: on) - new CMake option
AVIF_CODEC_DAV1D
to enable/disable the usage of dav1d's codec (default: off) codec_dav1d.c
, which provides decoding via libdav1d
- fuzz.sh which builds with afl-clang and runs afl-fuzz
- aviffuzz tool, used in fuzzing script
- fuzz inputs made with colorist
.clang-format
fileavifArray*()
functions for basic dynamic arrays when parsingmoov
box parsing- now reads ‘avis’ brands
- Split avifDecoderRead() into components for image sequences:
- avifDecoderSetSource()
- avifDecoderParse()
- avifDecoderNextImage()
- avifImageCopy()
- avifDecoderReset()
- Added decoder and image timings for image sequences
Changed
- Reorganized internal struct avifCodec to accommodate multiple codecs simultaneously (compile time; not exposed to API)
- Fix some compiler warnings
- Sanity check offsets and sizes in items table before using
- Bail out of box header advertises an impossible size
- Ran clang-format on all of src and include
- Fix copypasta leading to a memory leak in RGB planes
- Switched items and properties during parse to use dynamic arrays
- Refactored codec API to not require each codec to maintain per-plane decoder instances
- avifImage can now “not own” its planes and directly point at decoder planes to avoid copies
- aviffuzz attempts to decode all images in source material twice (using avifDecoderReset())
- Switch decoder->quality to explicit [minQuantizer, maxQuantizer], update assoc. constants
- Add examples to README
0.2.0 - 2019-06-12
Added
- Added
avifEncoder
and avifDecoder
to match avifImage
's pattern and allow for easier future parameterization
Changed
- Renamed project in cmake to
libavif
to match new official repo naming - Updated appveyor script to use
libavif
- Updated examples and apps to use new encoder/decoder pattern
0.1.4 - 2019-06-11
Added
avifPixelFormatToString()
convenience function for debugging/printingavifenc
and avifdec
“apps” which show basic bidirectional conversion to y4m
Changed
- Make calling
avifImageYUVToRGB()
upon reading an avif optional - Moved
ext/aom
submodule to use official remote - Update
ext/aom
submodule to commit 38711e7fe
Removed
- Remove all calls to
convertXYZToXYY()
as they were all unnecessary
0.1.3 - 2019-04-23
Changed
ftyp
- Change major_brand
to avif
ftyp
- Reorder compatible_brands
, add MA1A
or MA1B
when appropriate- Write
meta
box before mdat
box for streaming friendliness
0.1.2 - 2019-04-18
Added
AVIF_NCLX_COLOUR_PRIMARIES_P3
(convenient mirrored value)avifNclxColourPrimariesFind()
- Finds a builtin avifNclxColourPrimaries and name by a set of primaries
Changed
- Fixed enum name copypasta for
AVIF_NCLX_COLOUR_PRIMARIES_EG432_1
- Fix UV limited ranges when doing full<->limited range conversion
0.1.1 - 2019-04-15
Added
- Added
appveyor.yml
(exported from Appveyor) - Move
ext/aom
to a proper submodule - Update AOM to commit 3e3b9342a with minor CMake changes
Changed
- Added static library artifact zip to Windows x64 builds (Appveyor)
- Updated README to explain libavif's goals and a little more build info
- Fix clang warning in
avifVersion()
signature
0.1.0 - 2019-04-12
Added
- First version. Plenty of bugfixes and features await!
ext/aom
based off AOM commit 3563b12b with minor CMake changes- An interest and willingness to maintain this file.
- Constants
AVIF_VERSION
, AVIF_VERSION_MAJOR
, AVIF_VERSION_MINOR
, AVIF_VERSION_PATCH
avifVersion()
function