Bitstream Semantics

This section specifies the meaning of the syntax elements read in the syntax structures.

OBU Semantics

The bitstream consists of a sequence of open bitstream units (OBUs).

Each OBU is given to the decoding process in turn as a bitstream along with a variable sz that gives the total number of bytes in the OBU.

Note: Encoders should always set sz equal to the actual size of the OBU. Decoders should ignore any extra bytes if sz is greater than the actual size of the OBU. (For example, this may be needed if a future version of AV1 defines some additional fields at the end of the frame header, that are not strictly required for decoding, but if interpreted can improve the decoding). {:.alert .alert-info }

OBUs must begin at a byte boundary.

Methods of framing the OBUs in a container format are outside the scope of this Specification. However, one simple method of packing OBUs is described in Annex B.

OBU Header Semantics

OBUs are structured with a header and a payload. The header identifies the type of the payload using the obu_type header parameter.

obu_forbidden_bit must be set to 0.

Note: This ensures that MPEG2 transport is possible by preventing emulation of MPEG2 transport stream ids. {:.alert .alert-info }

obu_type specifies the type of data structure contained in the OBU payload:

obu_type	Name of obu_type
0	Reserved
1	OBU_SEQUENCE_HEADER
2	OBU_TD
3	OBU_FRAME_HEADER
4	OBU_TILE_GROUP
5	OBU_METADATA
6-14	Reserved
15	OBU_PADDING
{:.table .table-sm .table-bordered }

Reserved units are for future use and shall be ignored by AV1 decoder.

obu_reserved_2bits must be set to 0. The value is ignored by a decoder.

obu_extension_flag indicates if the optional obu_extension_header is present.

OBU Extension Header Semantics

temporal_id specifies the temporal level of the data contained in the OBU.

enhancement_id specifies the enhancement level of the data contained in the OBU.

Tile group OBU data associated with enhancement_id and temporal_id equal to 0 are referred to as the base layer, whereas tile group OBU data that are associated with enhancement_id greater than 0 or temporal_id greater than 0 are referred to as enhancement layer(s).

Coded video data of a temporal level with temporal_id T and enhancement level with enhancement_id S are only allowed to reference previously coded video data of temporal_id T' and enhancement_id S', where T' <= T and S' <= S.

extension_header_reserved_3bits must be set to 0. The value is ignored by a decoder.

Trailing Bits Semantics

zero_bit shall be equal to 0 and is inserted into the bitstream to align the bit position to a multiple of 8 bits.

Reserved OBU Semantics

The reserved OBU allows the extension of this specification with additional OBU types in a way that allows older decoders to ignore them.

reserved_obu_payload_byte is a payload byte of the reserved OBU.

Sequence Header OBU Semantics

profile_low_bit and profile_high_bit combine to make the variable Profile. AV1 supports 3 profiles:

Profile	Bit depth	SRGB Colorspace support	Chroma subsampling
0	8 or 10	No	YUV 4:2:0
1	8 or 10	Yes	YUV 4:4:4
2	8 or 10	No	YUV 4:2:2
2	12	Yes	YUV 4:2:0, YUV 4:2:2, YUV 4:4:4
{:.table .table-sm .table-bordered }

Note: It is a requirement of bitstream conformance that profile_low_bit and profile_high_bit are never set at the same time, i.e. that Profile is not equal to 3. {:.alert .alert-info }

reserved_zero shall be equal to 0.

level[ i ] specifies the level that the coded video sequence conforms to when all layers up to and including i are included in the decoding process.

enhancement_layers_cnt indicates the maximum number of enhancement layers that may be present in the sequence.

frame_width_bits_minus_1 specifies the number of bits minus 1 used for transmitting the frame width syntax elements.

frame_height_bits_minus_1 specifies the number of bits minus 1 used for transmitting the frame height syntax elements.

max_frame_width_minus_1 specifies the maximum frame width minus 1 for the frames represented by this sequence header.

max_frame_height_minus_1 specifies the maximum frame height minus 1 for the frames represented by this sequence header.

frame_id_numbers_present_flag specifies whether frame id numbers are present in the bitstream.

Note: Frame id numbers are not needed by the decoding process, but allow decoders to spot when frames have been missed and take an appropriate action. {:.alert .alert-info }

frame_id_length_minus1 is used to calculate the number of bits used to encode the frame_id syntax element.

delta_frame_id_length_minus2 specifies the number of bits minus 2 used to encode delta_frame_id syntax elements.

timing_info_present_flag specifies whether timing info is present in the bitstream.

film_grain_params_present specifies whether film grain parameters are present in the bitstream.

Color Config Semantics

high_bitdepth and twelve_bit are syntax elements which, together with the variable Profile, determine the bit depth.

monochrome equal to 1 indicates that the video does not contain U and V color planes. monochrome equal to 0 indicates that the video contains Y, U, and V color planes.

color_description_present_flag equal to 1 specifies that color_primaries, transfer_characteristics, and matrix_coeffs are present. color_description_present_flag equal to 0 specifies that color_primaries, transfer_characteristics and matrix_coeffs are not present.

color_primaries is an integer that is defined by the "Color primaries" section of ISO/IEC 23001-8:2016.

color_primaries	Name of color primaries	Description
1	CP_BT_709	BT.709
2	CP_UNSPECIFIED	Unspecified
4	CP_BT_470_M	BT.470 System M (historical)
5	CP_BT_470_B_G	BT.470 System B, G (historical)
6	CP_BT_601	BT.601
7	CP_SMPTE_240	SMPTE 240
8	CP_GENERIC_FILM	Generic film (color filters using illuminant C)
9	CP_BT_2020	BT.2020, BT.2100
10	CP_XYZ	SMPTE 428 (CIE 1921 XYZ)
11	CP_SMPTE_431	SMPTE RP 431-2
12	CP_SMPTE_432	SMPTE EG 432-1
22	CP_EBU_3213	EBU Tech. 3213-E
{:.table .table-sm .table-bordered }

transfer_characteristics is an integer that is defined by the "Transfer characteristics" section of ISO/IEC 23001-8:2016.

transfer_characteristics	Name of transfer characteristics	Description
0	TC_RESERVED_0	For future use
1	TC_BT_709	BT.709
2	TC_UNSPECIFIED	Unspecified
3	TC_RESERVED_3	For future use
4	TC_BT_470_M	BT.470 System M (historical)
5	TC_BT_470_B_G	BT.470 System B, G (historical)
6	TC_BT_601	BT.601
7	TC_SMPTE_240	SMPTE 240 M
8	TC_LINEAR	Linear
9	TC_LOG_100	Logarithmic (100 : 1 range)
10	TC_LOG_100_SQRT10	Logarithmic (100 * Sqrt(10) : 1 range)
11	TC_IEC_61966	IEC 61966-2-4
12	TC_BT_1361	BT.1361
13	TC_SRGB	sRGB or sYCC
14	TC_BT_2020_10_BIT	BT.2020 10-bit systems
15	TC_BT_2020_12_BIT	BT.2020 12-bit systems
16	TC_SMPTE_2084	SMPTE ST 2084, ITU BT.2100 PQ
17	TC_SMPTE_428	SMPTE ST 428
18	TC_HLG	BT.2100 HLG, ARIB STD-B67
{:.table .table-sm .table-bordered }

matrix_coefficients is an integer that is defined by the "Matrix coefficients" section of ISO/IEC 23001-8:2016.

matrix_coefficients	Name of matrix coefficients	Description
0	MC_IDENTITY	Identity matrix
1	MC_BT_709	BT.709
2	MC_UNSPECIFIED	Unspecified
3	MC_RESERVED_3	For future use
4	MC_FCC	US FCC 73.628
5	MC_BT_470_B_G	BT.470 System B, G (historical)
6	MC_BT_601	BT.601
7	MC_SMPTE_240	SMPTE 240 M
8	MC_SMPTE_YCGCO	YCgCo
9	MC_BT_2020_NCL	BT.2020 non-constant luminance, BT.2100 YCbCr
10	MC_BT_2020_CL	BT.2020 constant luminance
11	MC_SMPTE_2085	SMPTE ST 2085 YDzDx
12	MC_CHROMAT_NCL	Chromaticity-derived non-constant luminance
13	MC_CHROMAT_CL	Chromaticity-derived constant luminance
14	MC_ICTCP	BT.2100 ICtCp
{:.table .table-sm .table-bordered }

If (color_primaries == CP_BT_709 and transfer_characteristics == TC_SRGB and matrix_coefficients == MC_IDENTITY), it is a requirement of bitstream conformance that ( Profile == 1 || (Profile == 2 && BitDepth==12) ).

color_range specifies the black level and range of the luma and chroma signals as specified in Rec. ITU-R BT.709-6 and Rec. ITU-R BT.2020-2:

color_range	Description	Details
0	Studio swing	For BitDepth equals 8: Y is between 16 and 235 inclusive. U and V are between 16 and 240 inclusive.
		For BitDepth equals 10: Y is between 64 and 940 inclusive. U and V are between 64 and 960 inclusive.
		For BitDepth equals 12: Y is between 256 and 3760. U and V are between 256 and 3840 inclusive.
1	Full swing	No restriction on Y, U, V values.
{:.table .table-sm .table-bordered }

Note: Note that this specification does not enforce the range of YUV values when the YUV range is signaled as Studio swing. Therefore the application should perform additional clamping and color conversion operations according to the specified range. {:.alert .alert-info }

subsampling_x, subsampling_y specify the chroma subsampling format:

subsampling_x	subsampling_y	Description
0	0	YUV 4:4:4
0	1	YUV 4:4:0
1	0	YUV 4:2:2
1	1	YUV 4:2:0
{:.table .table-sm .table-bordered }

chroma_sample_position specifies the sample position for subsampled streams:

| chroma_sample_position | Name of chroma sample position| Description |:----------------------:|:------------------------------: ----------- | 0 | CSP_UNKNOWN | Unknown (in this case the source video transfer function must be signaled outside the AV1 bitstream) | 1 | CSP_VERTICAL | Horizontally co-located with (0, 0) luma sample, vertical position in the middle between two luma samples | 2 | CSP_COLOCATED | co-located with (0, 0) luma sample | 3 | CSP_RESERVED | {:.table .table-sm .table-bordered }

separate_uv_delta_q equal to 1 indicates that the U and V planes may have separate delta quantizer values. separate_uv_delta_q equal to 0 indicates that the U and V planes will share the same delta quantizer value.

Timing Info Semantics

num_units_in_tick is the number of time units of a clock operating at the frequency time_scale Hz that corresponds to one increment of a clock tick counter. A clock tick, in seconds, is equal to num_units_in_tick divided by time_scale.

It is a requirement of bitstream conformance that num_units_in_tick is greater than 0.

time_scale is the number of time units that pass in one second.

It is a requirement of bitstream conformance that time_scale is greater than 0.

equal_picture_interval equal to 1 indicates that pictures should be displayed according to their output order with the number of ticks between two consecutive pictures (without dropping frames) specified by num_ticks_per_picture_minus1 + 1. equal_picture_interval equal to 0 indicates that the interval between two consecutive pictures is not specified.

num_ticks_per_picture_minus1 plus 1 specifies the number of clock ticks corresponding to output time between two consecutive pictures in the output order.

It is a requirement of bitstream conformance that the value of num_ticks_per_picture_minus1 shall be in the range of 0 to (1 << 32) − 2, inclusive.

Note: The frame rate, when specified explicitly, applies to the top temporal layer of the bitstream. If bitstream is expected to be manipulated, e.g. by intermediate network elements, then the resulting frame rate may not match the specified one. In this case, an encoder is advised to use explicit time codes or some mechanisms that convey picture timing information outside the bitstream. {:.alert .alert-info }

Temporal Delimiter OBU Semantics

SeenFrameHeader is a variable used to mark whether we have received the frame header for the current frame.

Padding OBU Semantics

The padding OBU offers padding from 1 to 256 bytes. Multiple padding units can be present, each padding with an arbitrary number of bytes.

obu_padding_length indicates the length of the remaining payload in bytes.

obu_padding_byte is a padding bytes. Padding bytes may have arbitrary values and have no effect on the decoding process.

Metadata OBU Semantics

metadata_type indicates the type of metadata:

metadata_type	Name of metadata_type
0	METADATA_TYPE_PRIVATE_DATA
1	METADATA_TYPE_HDR_CLL
2	METADATA_TYPE_HDR_MDCV
3	METADATA_TYPE_SCALABILITY
4-65535	Reserved
{:.table .table-sm .table-bordered }

Metadata Private Data Semantics

Private data allows the transfer of arbitrary, application-specific binary information.

metadata_private_data_payload_byte[ i ] carries the payload of the private data. Private data may have arbitrary values and have no effect on the decoding process.

Metadata High Dynamic Range Content Light Level Semantics

max_cll specifies the maximum content light level as specified in CEA-861.3, Appendix A.

max_fall specifies the maximum frame-average light level as specified in CEA-861.3, Appendix A.

Metadata High Dynamic Range Mastering Display Color Volume Semantics

primary_chromaticity_x[ i ] specifies a 0.16 fixed-point X chromaticity coordinate as defined by CIE 1931, where i = 0,1,2 specifies Red, Green, Blue respectively.

primary_chromaticity_y[ i ] specifies a 0.16 fixed-point Y chromaticity coordinate as defined by CIE 1931, where i = 0,1,2 specifies Red, Green, Blue respectively.

white_point_chromaticity_x[ i ] specifies a 0.16 fixed-point white X chromaticity coordinate as defined by CIE 1931.

white_point_chromaticity_y[ i ] specifies a 0.16 fixed-point white Y chromaticity coordinate as defined by CIE 1931.

luminance_max is a 24.8 fixed-point maximum luminance, represented in candelas per square meter.

luminance_min is a 18.14 fixed-point minimum luminance, represented in candelas per square meter.

Metadata Scalability Semantics

scalability_mode_idc indicates the picture prediction structure of the bitstream.

scalability_mode_idc	Name of scalability_mode_idc
0	SCALABILITY_L1T2
1	SCALABILITY_L1T3
2	SCALABILITY_L2T1
3	SCALABILITY_L2T2
4	SCALABILITY_L2T3
5	SCALABILITY_S2T1
6	SCALABILITY_S2T2
7	SCALABILITY_S2T3
8	SCALABILITY_L2T2h
9	SCALABILITY_L2T3h
10	SCALABILITY_S2T1h
11	SCALABILITY_S2T2h
12	SCALABILITY_S2T3h
13	SCALABILITY_SS
14-255	reserved
{:.table .table-sm .table-bordered }

The scalability metadata provides two mechanisms for describing the underlying picture prediction structure of the bitstream:

Selection among a set of preconfigured structures, or modes, covering a number of cases that have found wide use in applications.
A facility for specifying picture prediction structures to accommodate a variety of special cases.

The preconfigured modes are described below. The mechanism for describing alternative structures is described in scalability_structure() below.

All predefined modes follow a dyadic, hierarchical picture prediction structure. They support 2 or 3 temporal layers, in combinations with zero or one enhancement layer. The enhancement layer may have twice or one and a half times the resolution of the base layer in each dimension, depending on the mode.
There is also support for an enhancement layer that uses no inter-layer prediction (i.e., the enhancement layer does not use its corresponding base layer as a reference).
The following table lists the predefined scalability structures.

Name of scalability_mode_idc	Spatial Layers	Resolution Ratio	Temporal Layers	Inter-layer dependency
SCALABILITY_L1T2	1		2
SCALABILITY_L1T3	1		3
SCALABILITY_L2T1	2	2:1	1	Yes
SCALABILITY_L2T2	2	2:1	2	Yes
SCALABILITY_L2T3	2	2:1	3	Yes
SCALABILITY_S2T1	2	2:1	1	No
SCALABILITY_S2T2	2	2:1	2	No
SCALABILITY_S2T3	2	2:1	3	No
SCALABILITY_L2T2h	2	1.5:1	2	Yes
SCALABILITY_L2T3h	2	1.5:1	3	Yes
SCALABILITY_S2T1h	2	1.5:1	1	No
SCALABILITY_S2T2h	2	1.5:1	2	No
SCALABILITY_S2T3h	2	1.5:1	3	No
{:.table .table-sm .table-bordered }

The following figures show the picture prediction structures for certain modes:

L1T2

L1T3

L2T1

L2T2

L2T3

S2T1

S2T2

S2T3

Scalability Structure Semantics

Note: The scalability_structure is intended for use by intermediate processing entities that may perform selective layer elimination. Its presence allows these entities to know the structure of the video bitstream without have to decode individual frames. Scalability structures should be placed immediately after the sequence header so that these entities are informed of the scalability structure of the video sequence as early as possible. {:.alert .alert-info }

enhancement_layers_cnt indicates the number of enhancement layers present in the video sequence.

enhancement_layer_description_present_flag indicates when set to 1 that the enhancement_layer_ref_id is present for each of the enhancement_layers_cnt layers, or that it is not present when set to 0.

enhancement_layer_dimensions_present_flag indicates when set to 1 that the enhancement_layer_max_width and enhancement_layer_max_height parameters are present for each of the enhancement_layers_cnt layers, or that it they are not present when set to 0.

temporal_group_description_present_flag indicates when set to 1 that the temporal dependency information is present, or that it is not when set to 0.

scalability_structure_reserved_3bits must be set to zero and be ignored by decoders.

enhancement_layer_max_width[ i ] specifies the maximum frame width for the frames with enhancement_id equal to i. This number must not be larger than max_frame_width_minus_1 + 1.

enhancement_layer_max_ height[ i ] specifies the maximum frame height for the frames with enhancement_id equal to i. This number must not be larger than max_frame_height_minus_1 + 1.

enhancement_layer_ref_id[ i ] specifies the enhancement_id value of the frame within the current temporal unit that the frame of layer i uses for reference. If no frame within the current temporal unit is used for reference the value must be equal to 255.

temporal_group_size indicates the number of pictures in a temporal picture group. If the temporal_group_size is greater than 0, then the scalability structure data allows the inter-picture temporal dependency structure of the video sequence to be specified. If the temporal_group_size is greater than 0, then for temporal_group_size pictures in the temporal group, each picture's temporal layer id (temporal_id), switch up point (temporal_group_switching_up_point_flag), and the reference picture indices (temporal_group_ref_pic_diff) are specified.

The first picture specified in a temporal group must have temporal_id equal to 0.

If the parameter temporal_group_size is not present or set to 0, then either there is only one temporal layer or there is no fixed inter-picture temporal dependency present going forward in the video sequence.

Note that for a given picture, all frames follow the same inter-picture temporal dependency structure.
However, the frame rate of each layer can be different from each other. The specified dependency structure in the scalability structure data must be for the highest frame rate layer.

temporal_group_temporal_id[ i ] specifies the temporal_id value for the i-th picture in the temporal group.

temporal_group_ref_cnt[ i ] indicates the number of reference pictures used by the i-th picture in the temporal group.

temporal_group_reserved_2bits[ i ] must be set to 0 and be ignored by decoders.

temporal_group_ref_pic_diff[ i ][ j ] indicates, for the i-th picture in the temporal group, the temporal distance between the i-th picture and the j-th reference picture used by the i-th picture. The temporal distance is measured in frames, counting only frames of identical enhancement_id values.

Note: The scalability structure description does not allow different temporal prediction structures across non-temporal layers (i.e., layers with different enhancement_id values). It also only allows for a single reference picture for inter-layer prediction. {:.alert .alert-info }

Frame Header OBU Semantics

It is a requirement of bitstream conformance that a sequence header OBU has been received before a frame header OBU.

frame_header_copy is a function call that indicates that a copy of the previous frame_header_obu should be inserted at this point.

Note: Bitstreams may contain several copies of the frame_header_obu interspersed with tile_group_obu to allow for greater error resilience. However, all copies must contain identical contents. {:.alert .alert-info }

TileNum is a variable giving the index (zero-based) of the current tile.

MaxTileSize is a variable that tracks the largest size in bytes of each tile. (This is needed because the CDFs are updated based on the largest tile decoded.)

decode_frame is a function call that indicates that the decode frame process specified in [section 7.2][] should be invoked.

Uncompressed Header Semantics

show_existing_frame equal to 1, indicates the frame indexed by frame_to_show_map_idx is to be displayed; show_existing_frame equal to 0 indicates that further processing is required.

frame_to_show_map_idx specifies the frame to be displayed. It is only available if show_existing_frame is 1.

display_frame_id provides the frame number for the frame to display. It is a requirement of bitstream conformance that whenever display_frame_id is read, the value matches the value of current_frame_id at the time that the frame indexed by frame_to_show_map_idx was stored.

frame_type specifies the type of the frame:

frame_type	Name of frame_type
0	KEY_FRAME
1	INTER_FRAME
2	INTRA_ONLY_FRAME
3	SWITCH_FRAME
{:.table .table-sm .table-bordered }

It is possible for bitstreams to start with a non key frame and still be decodable. In this case there are a number of additional constraints on the bitstream that are detailed in [section 7.3][].

show_frame equal to 1 specifies that this frame should be immediately output once decoded. show_frame equal to 0 specifies that this frame should not be immediately output. (It may be output later if a later uncompressed header uses show_existing_frame equal to 1).

error_resilient_mode equal to 1 indicates that error resilient mode is enabled; error_resilient_mode equal to 0 indicates that error resilient mode is disabled.

Note: Error resilient mode allows the syntax of a frame to be decoded independently of previous frames. {:.alert .alert-info }

current_frame_id specifies the frame number for the current frame. Frame numbers are additional information that do not affect the decoding process, but provide decoders with a way of detecting missing reference frames so that appropriate action can be taken.

frame_size_override_flag equal to 1 specifies that the frame size will be contained in the bitstream. frame_size_override_flag equal to 0 specifies that the frame size is equal to the size in the sequence header.

use_128x128_superblock, when equal to 1, indicates that superblocks are 128x128 pixels. When equal to 0, it indicates that superblocks are 64x64 pixels.

allow_screen_content_tools equal to 1 indicates that intra blocks may use palette encoding; allow_screen_content_tools equal to 0 indicates that palette encoding is never used.

allow_intrabc equal to 1 indicates that intra block copy may be used in this frame. allow_intrabc equal to 0 indicates that intra block copy is not allowed in this frame.

Note: intra block copy is only allowed in intra frames, and disables all loop filtering. force_integer_mv will be equal to 1 for intra frames, so only integer offsets are allowed in block copy mode. {:.alert .alert-info }

seq_choose_integer_mv equal to 1 indicates that the force_integer_mv syntax element will be present in the frame headers of inter and intra-only frames.

seq_force_integer_mv equal to SELECT_INTEGER_MV indicates that the force_integer_mv syntax element will be present in the frame headers of inter and intra-only frames. Otherwise, seq_force_integer_mv contains the value for force_integer_mv.

refresh_frame_flags contains a bitmask that specifies which reference frame slots will be updated with the current frame after it is decoded.

See [section 7.16][] for details of the frame update process.

ref_frame_idx specifies which reference frames are used by inter frames. It is a requirement of bitstream conformance that the selected reference frames match the current frame in bit depth, profile, chroma subsampling, and color space.

RefFrameSignBias specifies the intended direction of the motion vector in time for each reference frame. A sign bias equal to 0 indicates that the reference frame is a forwards reference (i.e. the reference frame is expected to be output before the current frame); a sign bias equal to 1 indicates that the reference frame is a backwards reference.

Note: The sign bias is just an indication that can improve the accuracy of motion vector prediction and is not constrained to reflect the actual output order of pictures. {:.alert .alert-info }

delta_frame_id_minus1 is used to calculate DeltaFrameId.

DeltaFrameId specifies the distance to the frame id for the reference frame.

RefFrameId specifies the frame id for the reference frame. It is a requirement of bitstream conformance that whenever RefFrameId is calculated, the value matches the value of current_frame_id at the time that the frame indexed by ref_frame_idx was stored.

force_integer_mv equal to 1 specifies that motion vectors will always be integers. force_integer_mv equal to 0 specifies that motion vectors can contain fractional bits.

allow_high_precision_mv equal to 0 specifies that motion vectors are specified to quarter pel precision; allow_high_precision_mv equal to 1 specifies that motion vectors are specified to eighth pel precision.

can_use_previous equal to 1 specifies that information (motion vectors and global motion settings) from a previous frame can be used when decoding the current frame. can_use_previous equal to 0 specifies that this information will not be used.

frame_parallel_decoding_mode equal to 1 indicates that parallel decoding mode is enabled; frame_parallel_decoding_mode equal to 0 indicates that parallel decoding mode is disabled.

Note: Parallel decoding mode means that the CDFs in the frame context are not updated at the end of the frame. This means that the next frame can start to be decoded as soon as the frame headers of the current frame have been processed. {:.alert .alert-info }

frame_offset_update provides a hint to the motion vector prediction as to how many frames later this frame is expected to be shown.

OrderHint specifies the offset of when this frame is expected to be shown relative to the last key frame.

Note: There is no requirement that OrderHint should reflect the true output order. As a guideline, the motion vector prediction is expected to be more accurate if the true output order is used for frames that will be shown later. If a frame is never to be shown (e.g. it has been constructed as an average of several frames for reference purposes), the encoder is free to choose whichever value of frame_offset_update will give the best compression. {:.alert .alert-info }

OrderHints specifies the expected output order for each reference frame.

CurrentVideoFrame counts the number of frames that have been shown since the last key frame. This variable should be initialized to zero in case the sequence starts with an intra only frame.

Note: The starting value of CurrentVideoFrame is unimportant because the decoding process only depends on the relative value between values of OrderHint. {:.alert .alert-info }

AllLossless is a variable that is equal to 1 when all segments use lossless encoding.

reduced_tx_set equal to 1 specifies that the frame is restricted to a reduced subset of the full set of transform types.

setup_past_independence is a function call that indicates that this frame can be decoded without dependence on previous coded frames. When this function is invoked the following takes place:

FeatureData[ i ][ j ] and FeatureEnabled[ i ][ j ] are set equal to 0 for i = 0..7 and j = 0..3.
PrevSegmentIds[ row ][ col ] is set equal to 0 for row = 0..MiRows-1 and col = 0..MiCols-1.
loop_filter_delta_enabled is set equal to 1.
loop_filter_ref_deltas[ INTRA_FRAME ] is set equal to 1.
loop_filter_ref_deltas[ LAST_FRAME ] is set equal to 0.
loop_filter_ref_deltas[ LAST2_FRAME ] is set equal to 0.
loop_filter_ref_deltas[ LAST3_FRAME ] is set equal to 0.
loop_filter_ref_deltas[ BWDREF_FRAME ] is set equal to 0.
loop_filter_ref_deltas[ GOLDEN_FRAME ] is set equal to -1.
loop_filter_ref_deltas[ ALTREF_FRAME ] is set equal to -1.
loop_filter_ref_deltas[ ALTREF2_FRAME ] is set equal to -1.
loop_filter_mode_deltas[ i ] is set equal to 0 for i = 0..1.
The cumulative distribution function arrays are reset to default values as follows:
- YModeCdf is set to a copy of Default_Y_Mode_Cdf
- UVModeCflNotAllowedCdf is set to a copy of Default_Uv_Mode_Cfl_Not_Allowed_Cdf
- UVModeCflAllowedCdf is set to a copy of Default_Uv_Mode_Cfl_Allowed_Cdf
- AngleDeltaCdf is set to a copy of Default_Angle_Delta_Cdf
- IntrabcCdf is set to a copy of Default_Intrabc_Cdf
- PartitionW8Cdf is set to a copy of Default_Partition_W8_Cdf
- PartitionW16Cdf is set to a copy of Default_Partition_W16_Cdf
- PartitionW32Cdf is set to a copy of Default_Partition_W32_Cdf
- PartitionW64Cdf is set to a copy of Default_Partition_W64_Cdf
- PartitionW128Cdf is set to a copy of Default_Partition_W128_Cdf
- SegmentIdCdf is set to a copy of Default_Segment_Id_Cdf
- SegmentIdPredictedCdf is set to a copy of Default_Segment_Id_Predicted_Cdf
- Tx8x8Cdf is set to a copy of Default_Tx_8x8_Cdf
- Tx16x16Cdf is set to a copy of Default_Tx_16x16_Cdf
- Tx32x32Cdf is set to a copy of Default_Tx_32x32_Cdf
- Tx64x64Cdf is set to a copy of Default_Tx_64x64_Cdf
- TxfmSplitCdf is set to a copy of Default_Txfm_Split_Cdf
- FilterIntraModeCdf is set to a copy of Default_Filter_Intra_Mode_Cdf
- FilterIntraCdf is set to a copy of Default_Filter_Intra_Cdf
- InterpFilterCdf is set to a copy of Default_Interp_Filter_Cdf
- MotionModeCdf is set to a copy of Default_Motion_Mode_Cdf
- NewMvCdf is set to a copy of Default_New_Mv_Cdf
- ZeroMvCdf is set to a copy of Default_Zero_Mv_Cdf
- RefMvCdf is set to a copy of Default_Ref_Mv_Cdf
- CompoundModeCdf is set to a copy of Default_Compound_Mode_Cdf
- DrlModeCdf is set to a copy of Default_Drl_Mode_Cdf
- IsInterCdf is set to a copy of Default_Is_Inter_Cdf
- CompModeCdf is set to a copy of Default_Comp_Mode_Cdf
- SkipModeCdf is set to a copy of Default_Skip_Mode_Cdf
- SkipCdf is set to a copy of Default_Skip_Cdf
- CompRefCdf is set to a copy of Default_Comp_Ref_Cdf
- CompBwdRefCdf is set to a copy of Default_Comp_Bwd_Ref_Cdf
- SingleRefCdf is set to a copy of Default_Single_Ref_Cdf
- MvJointCdf[ i ] is set to a copy of Default_Mv_Joint_Cdf for i = 0..MV_CONTEXTS-1
- MvClassCdf[ i ] is set to a copy of Default_Mv_Class_Cdf for i = 0..MV_CONTEXTS-1
- MvClass0BitCdf[ i ][ comp ] is set to a copy of Default_Mv_Class0_Bit_Cdf for i = 0..MV_CONTEXTS-1 and comp = 0..1
- MvFrCdf[ i ] is set to a copy of Default_Mv_Fr_Cdf for i = 0..MV_CONTEXTS-1
- MvClass0FrCdf[ i ] is set to a copy of Default_Mv_Class0_Fr_Cdf for i = 0..MV_CONTEXTS-1
- MvClass0HpCdf[ i ][ comp ] is set to a copy of Default_Mv_Class0_Hp_Cdf for i = 0..MV_CONTEXTS-1 and comp = 0..1
- MvSignCdf[ i ][ comp ] is set to a copy of Default_Mv_Sign_Cdf for i = 0..MV_CONTEXTS-1 and comp = 0..1
- MvBitCdf[ i ][ comp ] is set to a copy of Default_Mv_Bit_Cdf for i = 0..MV_CONTEXTS-1 and comp = 0..1
- MvHpCdf[ i ][ comp ] is set to a copy of Default_Mv_Hp_Cdf for i = 0..MV_CONTEXTS-1 and comp = 0..1
- PaletteYModeCdf is set to a copy of Default_Palette_Y_Mode_Cdf
- PaletteUVModeCdf is set to a copy of Default_Palette_Uv_Mode_Cdf
- PaletteYSizeCdf is set to a copy of Default_Palette_Y_Size_Cdf
- PaletteUVSizeCdf is set to a copy of Default_Palette_Uv_Size_Cdf
- PaletteSize2YColorCdf is set to a copy of Default_Palette_Size_2_Y_Color_Cdf
- PaletteSize2UVColorCdf is set to a copy of Default_Palette_Size_2_Uv_Color_Cdf
- PaletteSize3YColorCdf is set to a copy of Default_Palette_Size_3_Y_Color_Cdf
- PaletteSize3UVColorCdf is set to a copy of Default_Palette_Size_3_Uv_Color_Cdf
- PaletteSize4YColorCdf is set to a copy of Default_Palette_Size_4_Y_Color_Cdf
- PaletteSize4UVColorCdf is set to a copy of Default_Palette_Size_4_Uv_Color_Cdf
- PaletteSize5YColorCdf is set to a copy of Default_Palette_Size_5_Y_Color_Cdf
- PaletteSize5UVColorCdf is set to a copy of Default_Palette_Size_5_Uv_Color_Cdf
- PaletteSize6YColorCdf is set to a copy of Default_Palette_Size_6_Y_Color_Cdf
- PaletteSize6UVColorCdf is set to a copy of Default_Palette_Size_6_Uv_Color_Cdf
- PaletteSize7YColorCdf is set to a copy of Default_Palette_Size_7_Y_Color_Cdf
- PaletteSize7UVColorCdf is set to a copy of Default_Palette_Size_7_Uv_Color_Cdf
- PaletteSize8YColorCdf is set to a copy of Default_Palette_Size_8_Y_Color_Cdf
- PaletteSize8UVColorCdf is set to a copy of Default_Palette_Size_8_Uv_Color_Cdf
- DeltaQCdf is set to a copy of Default_Delta_Q_Cdf
- DeltaLFCdf is set to a copy of Default_Delta_Lf_Cdf
- DeltaLFMultiCdf[ i ] is set to a copy of Default_Delta_Lf_Cdf for i = 0..FRAME_LF_COUNT-1
- IntraTxTypeSet1Cdf is set to a copy of Default_Intra_Tx_Type_Set1_Cdf
- IntraTxTypeSet2Cdf is set to a copy of Default_Intra_Tx_Type_Set2_Cdf
- InterTxTypeSet1Cdf is set to a copy of Default_Inter_Tx_Type_Set1_Cdf
- InterTxTypeSet2Cdf is set to a copy of Default_Inter_Tx_Type_Set2_Cdf
- InterTxTypeSet3Cdf is set to a copy of Default_Inter_Tx_Type_Set3_Cdf
- UseObmcCdf is set to a copy of Default_Use_Obmc_Cdf
- InterIntraCdf is set to a copy of Default_Inter_Intra_Cdf
- CompRefTypeCdf is set to a copy of Default_Comp_Ref_Type_Cdf
- CflSignCdf is set to a copy of Default_Cfl_Sign_Cdf
- UniCompRefCdf is set to a copy of Default_Uni_Comp_Ref_Cdf
- WedgeInterIntraCdf is set to a copy of Default_Wedge_Inter_Intra_Cdf
- CompGroupIdxCdf is set to a copy of Default_Comp_Group_Idx_Cdf
- CompoundIdxCdf is set to a copy of Default_Compound_Idx_Cdf
- CompoundTypeCdf is set to a copy of Default_Compound_Type_Cdf
- InterIntraModeCdf is set to a copy of Default_Inter_Intra_Mode_Cdf
- CflAlphaCdf is set to a copy of Default_Cfl_Alpha_Cdf
- UseWienerCdf is set to a copy of Default_Use_Wiener_Cdf
- UseSgrprojCdf is set to a copy of Default_Use_Sgrproj_Cdf
- RestorationTypeCdf is set to a copy of Default_Restoration_Type_Cdf
- TxbSkipCdf is set to a copy of Default_Txb_Skip_Cdf.
- EobPt16Cdf is set to a copy of Default_Eob_Pt_16_Cdf.
- EobPt32Cdf is set to a copy of Default_Eob_Pt_32_Cdf.
- EobPt64Cdf is set to a copy of Default_Eob_Pt_64_Cdf.
- EobPt128Cdf is set to a copy of Default_Eob_Pt_128_Cdf.
- EobPt256Cdf is set to a copy of Default_Eob_Pt_256_Cdf.
- EobPt512Cdf is set to a copy of Default_Eob_Pt_512_Cdf.
- EobPt1024Cdf is set to a copy of Default_Eob_Pt_1024_Cdf.
- EobExtraCdf is set to a copy of Default_Eob_Extra_Cdf.
- DcSignCdf is set to a copy of Default_Dc_Sign_Cdf.
- CoeffBaseEobCdf is set to a copy of Default_Coeff_Base_Eob_Cdf.
- CoeffBaseCdf is set to a copy of Default_Coeff_Base_Cdf.
- CoeffBrCdf is set to a copy of Default_Coeff_Br_Cdf.

load_cdfs( ctx ) is a function call that indicates that the CDF tables are loaded from frame context number ctx in the range 0 to (NUM_REF_FRAMES - 1). When this function is invoked, a copy of each CDF array mentioned in the semantics for setup_past_independence is loaded from an area of memory indexed by ctx. (The memory contents of these frame contexts have been initialized by previous calls to save_cdfs). Once the CDF arrays have been loaded, the last entry in each array is set to 0.

load_previous( ) is a function call that indicates that information from the previous frame should be loaded for use in decoding the current frame. When this function is invoked the following ordered steps apply:

The variable prevFrame is set equal to ref_frame_idx[ 0 ].
gm_params is set equal to SavedGmParams[ prevFrame ].
If segmentation_enabled is equal to 1, PrevSegmentIds[ row ][ col ] is set equal to SavedSegmentIds[ prevFrame ][ row ][ col ] for row = 0..MiRows-1, for col = 0..MiCols-1
If can_use_previous is equal to 1, the motion field estimation process in [section 7.6][] is invoked.

It is a requirement of bitstream conformance that if segmentation_enabled is equal to 1, prevFrame refers to a frame of the same width and height as the current frame. In other words, when this function is invoked, all of the following must be true:

RefMiCols[ prevFrame ] is equal to MiCols.
RefMiRows[ prevFrame ] is equal to MiRows.

Frame Size Semantics

frame_width_minus_1 plus one is the width of the frame in pixels.

frame_height_minus_1 plus one is the height of the frame in pixels.

It is a requirement of bitstream conformance that frame_width_minus_1 is less than or equal to max_frame_width_minus_1.

It is a requirement of bitstream conformance that frame_height_minus_1 is less than or equal to max_frame_height_minus_1.

Render Size Semantics

The render size is provided as a hint to the application about the desired display size. It has no effect on the decoding process.

render_and_frame_size_different equal to 0 means that the render width and height are inferred from the frame width and height. render_and_frame_size_different equal to 1 means that the render width and height are explicitly coded in the bitstream.

Note: It is legal for the bitstream to explicitly code the render dimensions in the bitstream even if they are an exact match for the frame dimensions. {:.alert .alert-info }

render_width_minus_1 plus one is the render width of the frame in pixels.

render_height_minus_1 plus one is the render height of the frame in pixels.

Frame Size with Refs Semantics

For inter frames, the frame size is either set equal to the size of a reference frame, or can be sent explicitly.

found_ref equal to 1 indicates that the frame dimensions can be inferred from reference frame i where i is the loop counter in the syntax parsing process for frame_size_with_refs. found_ref equal to 0 indicates that the frame dimensions are not inferred from reference frame i.

Once the FrameWidth and FrameHeight have been computed for an inter frame, it is a requirement of bitstream conformance that for all values of i in the range 0..(REFS_PER_FRAME - 1), all the following conditions are true:

2 * FrameWidth >= RefFrameWidth[ ref_frame_idx[ i ] ]
2 * FrameHeight >= RefFrameHeight[ ref_frame_idx[ i ] ]
FrameWidth <= 16 * RefFrameWidth[ ref_frame_idx[ i ] ]
FrameHeight <= 16 * RefFrameHeight[ ref_frame_idx[ i ] ]

Note: This is a requirement even if all the blocks in an inter frame are coded using intra prediction. {:.alert .alert-info }

Superres Params Semantics

use_superres equal to 0 indicates that no upscaling is needed. use_superres equal to 1 indicates that upscaling is needed.

coded_denom is used to compute the amount of upscaling.

SuperresDenom is the denominator of a fraction that specifies the ratio between the superblock width before and after upscaling. The numerator of this fraction is equal to the constant SUPERRES_NUM.

Compute Image Size Semantics

MiCols is the number of 4x4 block columns in the frame.

MiRows is the number of 4x4 block rows in the frame.

Interpolation Filter Semantics

is_filter_switchable equal to 1 indicates that the filter selection is signaled at the block level; is_filter_switchable equal to 0 indicates that the filter selection is signaled at the frame level.

interpolation_filter specifies the filter selection used for performing inter prediction:

interpolation_filter	Name of interpolation_filter
0	EIGHTTAP
1	EIGHTTAP_SMOOTH
2	EIGHTTAP_SHARP
3	BILINEAR
4	SWITCHABLE
{:.table .table-sm .table-bordered }

Loop Filter Semantics

loop_filter_level is an array containing loop filter strength values. Different loop filter strength values from the array are used depending on the image plane being filtered, and the edge direction (vertical or horizontal) being filtered.

loop_filter_sharpness indicates the sharpness level. The loop_filter_level and loop_filter_sharpness together determine when a block edge is filtered, and by how much the filtering can change the sample values.

The loop filter process is described in [section 7.11][].

loop_filter_delta_enabled equal to 1 means that the filter level depends on the mode and reference frame used to predict a block. loop_filter_delta_enabled equal to 0 means that the filter level does not depend on the mode and reference frame.

loop_filter_delta_update equal to 1 means that the bitstream contains additional syntax elements that specify which mode and reference frame deltas are to be updated. loop_filter_delta_update equal to 0 means that these syntax elements are not present.

update_ref_delta equal to 1 means that the bitstream contains the syntax element loop_filter_ref_delta; update_ref_delta equal to 0 means that the bitstream does not contain this syntax element.

loop_filter_ref_deltas contains the adjustment needed for the filter level based on the chosen reference frame. If this syntax element is not present in the bitstream, it maintains its previous value.

update_mode_delta equal to 1 means that the bitstream contains the syntax element loop_filter_mode_deltas; update_mode_delta equal to 0 means that the bitstream does not contain this syntax element.

loop_filter_mode_deltas contains the adjustment needed for the filter level based on the chosen mode. If this syntax element is not present in the bitstream, it maintains its previous value.

Note: The previous values for loop_filter_mode_deltas and loop_filter_ref_deltas are intially set by the setup_past_independence function and can be subsequently modified by these syntax elements being coded in a previous frame. {:.alert .alert-info }

Quantization Params Semantics

The residual is specified via decoded coefficients which are adjusted by one of four quantization parameters before the inverse transform is applied. The choice depends on the plane (Y or UV) and coefficient position (DC/AC coefficient). The Dequantization process is specified in [section 7.9][].

base_q_idx indicates the base frame qindex. This is used for Y AC coefficients and as the base value for the other quantizers.

DeltaQYDc indicates the Y DC quantizer relative to base_q_idx.

diff_uv_delta equal to 1 indicates that the U and V delta quantizer values are coded separately. diff_uv_delta equal to 0 indicates that the U and V delta quantizer values share a common value.

DeltaQUDc indicates the U DC quantizer relative to base_q_idx.

DeltaQUAc indicates the U AC quantizer relative to base_q_idx.

DeltaQVDc indicates the V DC quantizer relative to base_q_idx.

DeltaQVAc indicates the V AC quantizer relative to base_q_idx.

using_qmatrix specifies that the quantizer matrix will be used to compute quantizers.

min_qmlevel specifies the smallest level in the quantizer matrix that may be used.

max_qmlevel specifies the largest level in the quantizer matrix that may be used.

Delta Quantizer Semantics

delta_coded specifies that the delta_q syntax element is present in the bitstream.

delta_q specifies an offset (relative to base_q_idx) for a particular quantization parameter.

Segmentation Params Semantics

AV1 provides a means of segmenting the image and then applying various adjustments at the segment level.

Up to 8 segments may be specified for any given frame. For each of these segments it is possible to specify:

A quantizer (absolute value or delta).
A loop filter strength (absolute value or delta).
A prediction reference frame.
A block skip mode that implies both the use of a (0,0) motion vector and that no residual will be coded.

Each of these data values for each segment may be individually updated at the frame level. Where a value is not updated in a given frame, the value from the previous frame persists. The exceptions to this are key frames, intra only frames or other frames where independence from past frame values is required (for example to enable error resilience). In such cases all values are reset as described in the semantics for setup_past_independence.

The segment affiliation (the segmentation map) is stored at the resolution of 4x4 blocks. If no explicit update is coded for a block's segment affiliation, then it persists from frame to frame (until reset by a call to setup_past_independence).

SegIdPreSkip equal to 1 indicates that the segment id will be read before the skip syntax element. SegIdPreSkip equal to 0 indicates that the skip syntax element will be read first.

LastActiveSegId indicates the highest numbered segment id that has some enabled feature. This is used when decoding the segment id to only decode choices corresponding to used segments.

segmentation_enabled equal to 1 indicates that this frame makes use of the segmentation tool; segmentation_enabled equal to 0 indicates that the frame does not use segmentation.

segmentation_update_map equal to 1 indicates that the segmentation map are updated during the decoding of this frame. segmentation_update_map equal to 0 means that the segmentation map from the previous frame is used.

segmentation_temporal_update equal to 1 indicates that the updates to the segmentation map are coded relative to the existing segmentation map. segmentation_temporal_update equal to 0 indicates that the new segmentation map is coded without reference to the existing segmentation map.

segmentation_update_data equal to 1 indicates that new parameters are about to be specified for each segment. segmentation_update_data equal to 0 indicates that the segmentation parameters should keep their existing values.

feature_enabled equal to 0 indicates that the corresponding feature is unused and has value equal to 0. feature_enabled equal to 1 indicates that the feature value is coded in the bitstream.

feature_value specifies the feature data for a segment feature.

Tile Info Semantics

uniform_tile_spacing_flag equal to 1 means that the tiles are uniformly spaced across the frame. (In other words, all tiles are the same size except for the ones at the right and bottom edge which can be smaller.) uniform_tile_spacing_flag equal to 0 means that the tile sizes are coded.

increment_tile_cols_log2 indicates whether the tile width is increased.

TileColsLog2 specifies the base 2 logarithm of the desired number of tiles across the frame.

TileCols specifies the number of tiles across the frame. It is a requirement of bitstream conformance that TileCols is less than or equal to MAX_TILE_COLS.

increment_tile_rows_log2 is used to compute TileRowsLog2.

TileRowsLog2 specifies the base 2 logarithm of the desired number of tiles down the frame.

Note: For small frame sizes the actual number of tiles in the frame may be smaller than the desired number because the tile size is rounded up to a multiple of the maximum superblock size. {:.alert .alert-info }

TileRows specifies the number of tiles down the frame. It is a requirement of bitstream conformance that TileRows is less than or equal to MAX_TILE_ROWS.

sizeSb is used to specify the width/height of each tile in units of 128x128 pixels. It is a requirement of bitstream conformance that sizeSb is less than or equal to MAX_TILE_WIDTH_SB.

If uniform_tile_spacing_flag is equal to 0, it is a requirement of bitstream conformance that startSb is equal to sbCols when the loop writing MiColStarts exits.

If uniform_tile_spacing_flag is equal to 0, it is a requirement of bitstream conformance that startSb is equal to sbRows when the loop writing MiRowStarts exits.

Note: The requirements on startSb ensure that the sizes of each tile add up to the full size of the frame when measured in 128x128 units. {:.alert .alert-info }

MiColStarts is an array specifying the start column (in units of 4x4 pixels) for each tile across the image.

MiRowStarts is an array specifying the start row (in units of 4x4 pixels) for each tile down the image.

maxTileHeightSb specifies the maximum height (in units of 128x128 pixels) that can be used for a tile (to avoid making tiles with too much area).

maxTileHeightMi contains the maximum height (in units of 4x4 pixels) that can be used for a tile.

AllowDependentTileRow is an array that specifies which tile rows are allowed to use syntax dependencies between tiles.

Note: The AllowDependentTileRow array ensures that the syntax decode can be split into independent units of a certain maximum size even when dependent_tiles is equal to 1. {:.alert .alert-info }

dependent_tiles equal to 0 indicates that there are no syntax dependencies between tiles, dependent_tiles equal to 1 indicates that there can be dependencies between tiles.

Note: When dependent_tiles is equal to 1, the context determination, motion vector prediction, and intra prediction are allowed to depend on the tile above. {:.alert .alert-info }

loop_filter_across_tiles_v equal to 1 indicates that loop filtering is allowed across vertical tile edges. loop_filter_across_tiles_v equal to 0 indicates that loop filtering is not allowed across vertical tile edges.

loop_filter_across_tiles_h equal to 1 indicates that loop filtering is allowed across horizontal tile edges. loop_filter_across_tiles_h equal to 0 indicates that loop filtering is not allowed across horizontal tile edges.

tile_size_bytes_minus_1 is used to compute TileSizeBytes.

TileSizeBytes specifies the number of bytes needed to code each tile size.

Quantizer Index Delta Parameters Semantics

delta_q_present specifies whether quantizer index delta values are present in the bitstream.

delta_q_res specifies the left shift which should be applied to decoded quantizer index delta values.

Loop Filter Delta Parameters Semantics

delta_lf_present specifies whether loop filter delta values are present in the bitstream.

delta_lf_res specifies the left shift which should be applied to decoded loop filter delta values.

delta_lf_multi equal to 1 specifies that separate loop filter deltas are sent for horizontal luma edges, vertical luma edges, the U edges, and the V edges. delta_lf_multi equal to 0 specifies that the same loop filter delta is used for all edges.

Global Motion Params Semantics

is_global specifies whether global motion parameters are present for a particular reference frame.

is_rot_zoom specifies whether a particular reference frame uses rotation and zoom global motion.

is_translation specifies whether a particular reference frame uses translation global motion.

Global Param Semantics

absBits is used to compute the range of values that can be used for this parameter. The values allowed are in the range -(1 << absBits) to (1 << absBits).

precBits specifies the number of fractional bits used for representing this parameter in the bitstream. All global motion parameters are stored in the model with WARPEDMODEL_PREC_BITS fractional bits, but the parameters are encoded with less precision in the bitstream.

Decode Subexp Semantics

subexp_more_bits equal to 0 specifies that the parameter is in the range mk to mk+a-1. subexp_more_bits equal to 0 specifies that the parameter is greater than mk+a-1.

subexp_bits specifies the value of the parameter minus mk.

Decode Uniform Semantics

v and extra_bit are used to compute the value of an unsigned number in the range 0 to n-1.

Film Grain Params Semantics

apply_grain equal to 1 specifies that film grain should be added to this frame. apply_grain equal to 0 specifies that film grain should not be added.

grain_seed specifies the starting value for the pseudo-random numbers used during film grain synthesis.

update_grain equal to 1 means that a new set of parameters should be sent. update_grain equal to 0 means that the previous set of parameters should be used.

It is a requirement of bitstream conformance that if update_grain is set to 0, then there must have been a previous frame in decoding order for which update_grain was set to 1 (this ensures that the parameters are always defined before being used).

num_y_points specifies the number of coordinate pairs specified for the Y plane lookup.

It is a requirement of bitstream conformance that num_y_points is less than or equal to 14.

scaling_points_y[ i ][ 0 ] specifies the abscissa for coordinate i of the Y plane lookup.

scaling_points_y[ i ][ 1 ] specifies the ordinate for coordinate i of the Y plane lookup.

chroma_scaling_from_luma specifies that the chroma scaling is inferred from the luma scaling.

num_cb_points specifies the number of ordinate pairs specified for the U plane lookup.

It is a requirement of bitstream conformance that num_cb_points is less than or equal to 10.

scaling_points_cb[ i ][ 0 ] specifies the abscissa for coordinate i of the U plane lookup.

scaling_points_cb[ i ][ 1 ] specifies the ordinate for coordinate i of the U plane lookup.

num_cr_points specifies the number of ordinate pairs specified for the V plane lookup.

It is a requirement of bitstream conformance that num_cr_points is less than or equal to 10.

scaling_points_cr[ i ][ 0 ] specifies the abscissa for coordinate i of the V plane lookup.

scaling_points_cr[ i ][ 1 ] specifies the ordinate for coordinate i of the V plane lookup.

grain_scaling_minus_8 specifies a scaling shift used during film grain synthesis.

ar_coeffs_y_plus_128 specifies auto-regressive coefficients used for the Y plane.

ar_coeffs_cb_plus_128 specifies auto-regressive coefficients used for the U plane.

ar_coeffs_cr_plus_128 specifies auto-regressive coefficients used for the V plane.

ar_coeff_shift_minus_6 specifies a coefficient scaling shift used during film grain synthesis.

overlap_flag specifies if the filtering in film grain synthesis is allowed to access samples outside the current block.

clip_to_restricted_range equal to 1 specifies that the final sample values should be constrained to the studio swing ranges (see the semantics for color_range for an explanation of studio swing). clip_to_restricted_range equal to 0 specifies that the final sample values are allowed the full swing ranges.

TX Mode Semantics

tx_mode_select is used to compute TxMode.

TxMode specifies how the transform size is determined:

TxMode	Name of TxMode
0	ONLY_4X4
1	TX_MODE_LARGEST
2	TX_MODE_SELECT
{:.table .table-sm .table-bordered }

For tx_mode equal to TX_MODE_LARGEST, the inverse transform will use the largest transform size that fits inside the block.

For tx_mode equal to ONLY_4X4, the inverse transform will use only 4x4 transforms.

For tx_mode equal to TX_MODE_SELECT, the choice of transform size is specified explicitly for each block.

Skip Mode Params Semantics

ForwardFrame and BackwardFrame specify the frames to use for compound prediction when skip_mode is equal to 1.

skip_mode_present equal to 1 specifies that the syntax element skip_mode will be coded in the bitstream. skip_mode_present equal to 0 specifies that skip_mode will not be used for this frame.

Note: ForwardFrame and BackwardFrame are set to the closest forward and backward frames (as measured by frame_offset values). If no backward frame is found, then the second closest forward frame is used. (Forward prediction is when we use a frame for reference that is considered to be output before the current frame, backward prediction is when we use a frame that has not yet been output.) {:.alert .alert-info }

Frame Reference Mode Semantics

reference_select equal to 1 specifies that the mode info for inter blocks contains the syntax element comp_mode that indicates whether to use single or compound reference prediction. Reference_select equal to 0 specifies that all inter blocks will use single prediction.

allow_interintra_compound equal to 1 specifies that the mode info for inter blocks may contain the syntax element interintra. allow_interintra_compound equal to 0 specifies that the syntax element interintra will not be present in this frame.

allow_masked_compound equal to 1 specifies that the mode info for inter blocks may contain the syntax element compound_type. allow_masked_compound equal to 0 specifies that the syntax element compound_type will not be present in this frame.

Tile Group OBU Semantics

NumTiles specifies the total number of tiles in the frame.

tg_start specifies the zero-based index of the first tile in the current tile group.

It is a requirement of bitstream conformance that the value of tg_start is equal to the value of TileNum at the point that tile_group_obu is invoked.

tg_end specifies the zero-based index of the last tile in the current tile group.

It is a requirement of bitstream conformance that the value of tg_end is greater than or equal to tg_start.

It is a requirement of bitstream conformance that the value of tg_end for the last tile group in each frame is equal to TileNum - 1.

Note: These requirements ensure that conceptually all tile groups are present and received in order for the purposes of specifying the decode process. {:.alert .alert-info }

update_cdf is a function call that indicates that the CDF arrays are set equal to the final CDFs at the end of the largest tile (measured in bytes). This process is described in [section 7.5][].

tile_size specifies the size in bytes of the next coded tile.

Note: This size includes any padding bytes added by the exit process for the Boolean decoder. The size does not include the bytes used for tile_group_obu, trailing_bits, and tile_size. {:.alert .alert-info }

decode_frame is a function call that indicates that the decode frame process specified in [section 7.2][] should be invoked.

Decode Tile Semantics

clear_left_context is a function call that indicates that some arrays used to determine the probabilities are zeroed. When this function is invoked the arrays LeftLevelContext, LeftDcContext, and LeftSegPredContext are set equal to 0.

Note: LeftLevelContext[ plane ][ i ], LeftDcContext[ plane ][ i ], and LeftSegPredContext[ i ] need to be set to 0 for i = 0..MiRows-1, for plane = 0..2. {:.alert .alert-info }

clear_above_context is a function call that indicates that some arrays used to determine the probabilities are zeroed. When this function is invoked the arrays AboveLevelContext, AboveDcContext, and AboveSegPredContext are set equal to 0.

Note: AboveLevelContext[ plane ][ i ], AboveDcContext[ plane ][ i ], and AboveSegPredContext[ i ] need to be set to 0 for i = 0..MiCols-1, for plane = 0..2. {:.alert .alert-info }

ReadDeltas specifies whether the current block is the first one in the current superblock. Delta values for the quantizer index and loop filter are only read on the first block of a superblock.

Clear Block Decoded Flags Semantics

notLastColumn equal to 1 indicates that the current superblock is not in the rightmost column of the current tile. notLastColumn equal to 0 indicates that the current superblock is in the rightmost column of the current tile.

BlockDecoded is an array which stores one boolean value per 4x4 sample block per plane in the current superblock, plus a border of one 4x4 sample block on all sides of the superblock. Except for the borders, a value of 1 in BlockDecoded indicates that the corresponding 4x4 sample block has been decoded. The borders are used when computing above-right and below-left availability along the top and left edges of the superblock.

Decode Partition Semantics

partition specifies how a block is partitioned:

partition	Name of partition
0	PARTITION_NONE
1	PARTITION_HORZ
2	PARTITION_VERT
3	PARTITION_SPLIT
4	PARTITION_HORZ_A
5	PARTITION_HORZ_B
6	PARTITION_VERT_A
7	PARTITION_VERT_B
8	PARTITION_HORZ_4
9	PARTITION_VERT_4
{:.table .table-sm .table-bordered }

The variable subSize is computed from partition and indicates the size of the component blocks within this block:

subSize	Name of subSize
0	BLOCK_4X4
1	BLOCK_4X8
2	BLOCK_8X4
3	BLOCK_8X8
4	BLOCK_8X16
5	BLOCK_16X8
6	BLOCK_16X16
7	BLOCK_16X32
8	BLOCK_32X16
9	BLOCK_32X32
10	BLOCK_32X64
11	BLOCK_64X32
12	BLOCK_64X64
13	BLOCK_64X128
14	BLOCK_128X64
15	BLOCK_128X128
16	BLOCK_4X16
17	BLOCK_16X4
18	BLOCK_8X32
19	BLOCK_32X8
20	BLOCK_16X64
21	BLOCK_64X16
22	BLOCK_32X128
23	BLOCK_128X32
{:.table .table-sm .table-bordered }

The dimensions of these blocks are given in width, height order (e.g. BLOCK_8X16 corresponds to a block that is 8 pixels wide, and 16 pixels high).

It is a requirement of bitstream conformance that get_plane_residual_size( subSize, 1 ) is not equal to BLOCK_INVALID every time subSize is computed.

Note: This requirement prevents the UV blocks from being too tall or too wide (i.e. having aspect ratios outside the range 1:2 to 2:1). {:.alert .alert-info }

split_or_vert is used to compute partition for blocks when only split or vert partitions are legal because of overlap with the right hand edge of the frame.

split_or_horz is used to compute partition for blocks when only split or horz partitions are legal because of overlap with the bottom edge of the frame.

Decode Block Semantics

MiRow is a variable holding the vertical location of the block in units of 4x4 pixels.

MiCol is a variable holding the horizontal location of the block in units of 4x4 pixels.

MiSize is a variable holding the size of the block with values having the same interpretation for the variable subSize.

HasChroma is a variable that specifies whether chroma information is coded for this block.

Variable AvailU is equal to 0 if the information from the block above cannot be used; AvailU is equal to 1 if the information from the block above can be used.

Variable AvailL is equal to 0 if the information from the block to the left can not be used; AvailL is equal to 1 if the information from the block to the left can be used.

Note: Information from a block in a different tile can be used in some circumstances if the block is above, but not if the block is to the left. {:.alert .alert-info }

Intra Frame Mode Info Semantics

This syntax is used when coding an intra block within an intra frame.

use_intrabc equal to 1 specifies that intra block copy should be used for this block. use_intrabc equal to 0 specifies that intra block copy should not be used.

intra_frame_y_mode specifies the direction of intra prediction filtering:

intra_frame_y_mode	Name of intra_frame_y_mode
0	DC_PRED
1	V_PRED
2	H_PRED
3	D45_PRED
4	D135_PRED
5	D117_PRED
6	D153_PRED
7	D207_PRED
8	D63_PRED
9	SMOOTH_PRED
10	SMOOTH_V_PRED
11	SMOOTH_H_PRED
12	TM_PRED
13	UV_CFL_PRED
{:.table .table-sm .table-bordered }

uv_mode specifies the chrominance intra prediction mode using values with the same interpretation as in the semantics for intra_frame_y_mode.

Note: It is a requirement of bitstream conformance that, if uv_mode is equal to UV_CFL_PRED, Max( Block_Width[ MiSize ], Block_Height[ MiSize ] ) <= 32. {:.alert .alert-info }

Intra Segment ID Semantics

Lossless is a variable which, if equal to 1, indicates that the block is coded using a special 4x4 transform designed for encoding frames that are bit-identical with the original frames.

Read Segment ID Semantics

segment_id specifies which segment is associated with the current intra block being decoded. It first read from the stream, and then postprocessed based on the predicted segment id.

It is a requirement of bitstream conformance that the postprocessed value of segment_id (i.e. the value returned by neg_deinterleave) is in the range 0 to LastActiveSegId (inclusive of endpoints).

Inter Segment ID Semantics

seg_id_predicted equal to 1 specifies that the segment_id are taken from the segmentation map. seg_id_predicted equal to 0 specifies that the syntax element segment_id are parsed.

Note: It is legal for seg_id_predicted to be equal to 0 even if the value coded for the segment_id is equal to predictedSegmentId. {:.alert .alert-info }

Skip Mode Semantics

skip_mode equal to 1 indicates that this block will use some default settings (that correspond to compound prediction) and so most of the mode info is skipped. skip_mode equal to 0 indicates that the mode info is not skipped.

Skip Semantics

skip equal to 0 indicates that there may be some transform coefficients to read for this block; skip equal to 1 indicates that there are no transform coefficients.

Quantizer Index Delta Semantics

delta_q_abs specifies the absolute value of the quantizer index delta value being decoded. If delta_q_abs is equal to DELTA_Q_SMALL, the value is encoded using delta_q_rem_bits and delta_q_abs_bits.

delta_q_rem_bits and delta_q_abs_bits encode the absolute value of the quantizer index delta value being decoded, where the absolute value of the quantizer index delta value is of the form:

(1 << delta_q_rem_bits) + delta_q_abs_bits + 1

delta_q_sign_bit equal to 0 indicates that the quantizer index delta value is positive; delta_q_sign_bit equal to 1 indicates that the quantizer index delta value is negative.

Loop Filter Delta Semantics

delta_lf_abs specifies the absolute value of the loop filter delta value being decoded. If delta_lf_abs is equal to DELTA_LF_SMALL, the value is encoded using delta_lf_rem_bits and delta_lf_abs_bits.

delta_lf_rem_bits and delta_lf_abs_bits encode the absolute value of the loop filter delta value being decoded, where the absolute value of the loop filter delta value is of the form:

( 1 << ( delta_lf_rem_bits + 1 ) ) + delta_lf_abs_bits + 1

delta_lf_sign_bit equal to 0 indicates that the loop filter delta value is positive; delta_lf_sign_bit equal to 1 indicates that the loop filter delta value is negative.

CDEF Params Semantics

cdef_damping_minus_3 controls the amount of damping in the deringing filter.

cdef_bits specifies the number of bits needed to specify which CDEF filter to apply.

cdef_y_pri_strength and cdef_uv_pri_strength specify the strength of the primary filter.

cdef_y_sec_strength and cdef_uv_sec_strength specify the strength of the secondary filter.

Loop Restoration Params Semantics

lr_type is used to compute FrameRestorationType.

FrameRestorationType specifies the type of restoration used for each plane as follows:

lr_type	FrameRestorationType	Name of FrameRestorationType
0	0	RESTORE_NONE
1	3	RESTORE_SWITCHABLE
2	1	RESTORE_WIENER
3	2	RESTORE_SGRPROJ
{:.table .table-sm .table-bordered }

lr_unit_shift specifies if the luma restoration size should be halved.

lr_unit_extra_shift specifies if the luma restoration size should be halved again.

lr_uv_shift is only present for 4:2:0 formats and specifies if the chroma size should be half the luma size.

LoopRestorationSize[plane] specifies for size of loop restoration units in units of samples in the current plane.

TX Size Semantics

tx_depth is used to compute TxSize. tx_depth is inverted with respect to TxSize, i.e. it specifies how much smaller the transform size should be made than the largest possible transform size for the block.

TxSize specifies the transform size to be used for this block:

TxSize	Name of TxSize
0	TX_4X4
1	TX_8X8
2	TX_16X16
3	TX_32X32
4	TX_64X64
5	TX_4X8
6	TX_8X4
7	TX_8X16
8	TX_16X8
9	TX_16X32
10	TX_32X16
11	TX_32X64
12	TX_64X32
13	TX_4X16
14	TX_16X4
15	TX_8X32
16	TX_32X8
17	TX_16X64
18	TX_64X16
{:.table .table-sm .table-bordered }

Note: TxSize is determined for skipped intra blocks because TxSize controls the granularity of the intra prediction. {:.alert .alert-info }

Note: When TxSize is first computed, it will only take values corresponding to square sizes. Rectangular sizes are inferred if this square TxSize is larger than the block size. TxSize is then set to be the largest rectangular size that fits in the block. {:.alert .alert-info }

Inter TX Size Semantics

MinTxSize is a variable that contains a square transform size that would fit inside all the transforms within the block. In other words, if the block contains transforms of size W[i] by H[i], then MinTxSize represents a transform of size m by m, where m is equal to the smallest value in both W and H.

InterTxSizes is an array that holds the transform sizes within inter frames.

Note: TxSizes and InterTxSizes contain different values. All the values in TxSizes across a residual block will share the same value, while InterTxSizes can represent several different transform sizes within a residual block. {:.alert .alert-info }

Var TX Size Semantics

txfm_split equal to 1 specifies that the block should be split into smaller transform sizes. txfm_split equal to 0 specifies that the block should not be split any more.

Transform Type Semantics

set specifies the transform set.

is_inter	set	Name of transform set
Don't care	0	TX_SET_DCTONLY
0	1	TX_SET_INTRA_1
0	2	TX_SET_INTRA_2
1	1	TX_SET_INTER_1
1	2	TX_SET_INTER_2
1	3	TX_SET_INTER_3
{:.table .table-sm .table-bordered }

The transform sets determine what subset of transform types can be used, according to the following table.

Transform type	TX_SET_DCTONLY	TX_SET_INTRA_1	TX_SET_INTRA_2	TX_SET_INTER_1	TX_SET_INTER_2	TX_SET_INTER_3
DCT_DCT	X	X	X	X	X	X
ADST_DCT		X	X	X	X
DCT_ADST		X	X	X	X
ADST_ADST		X	X	X	X
FLIPADST_DCT				X	X
DCT_FLIPADST				X	X
FLIPADST_FLIPADST				X	X
ADST_FLIPADST				X	X
FLIPADST_ADST				X	X
IDTX		X	X	X	X	X
V_DCT		X		X	X
H_DCT		X		X	X
V_ADST				X
H_ADST				X
V_FLIPADST				X
H_FLIPADST				X
{:.table .table-sm .table-bordered }

inter_tx_type specifies the transform type for inter blocks.

intra_tx_type specifies the transform type for intra blocks.

Is Inter Semantics

is_inter equal to 0 specifies that the block is an intra block; is_inter equal to 1 specifies that the block is an inter block.

Intra Block Mode Info Semantics

This syntax is used when coding an intra block within an inter frame.

y_mode and uv_mode specify the direction of intra prediction using values with the same interpretation as for intra_frame_y_mode.

Note: It is a requirement of bitstream conformance that, if uv_mode is equal to UV_CFL_PRED, Max( Block_Width[ MiSize ], Block_Height[ MiSize ] ) <= 32. {:.alert .alert-info }

Inter Block Mode Info Semantics

This syntax is used when coding an inter block.

compound_mode specifies how the motion vector used by inter prediction is obtained when using compound prediction. An offset is added to compound_mode to compute YMode as follows:

YMode	Name of YMode
14	NEARESTMV
15	NEARMV
16	GLOBALMV
17	NEWMV
18	NEAREST_NEARESTMV
19	NEAR_NEARMV
20	NEAREST_NEWMV
21	NEW_NEARESTMV
22	NEAR_NEWMV
23	NEW_NEARMV
24	GLOBAL_GLOBALMV
25	NEW_NEWMV
{:.table .table-sm .table-bordered }

Note: The intra modes take values 0..13 so these YMode values start at 14. {:.alert .alert-info }

new_mv equal to 0 means that a motion vector difference should be read.

zero_mv equal to 0 means that the motion vector should be set equal to default motion for the frame.

ref_mv equal to 0 means that the most likely motion vector should be used (called NEAREST), ref_mv equal to 1 means that the second most likely motion vector should be used (called NEAR).

interp_filter specifies the type of filter used in inter prediction. Values 0..3 are allowed with the same interpretation as for interpolation_filter. One filter type is specified for the vertical filter direction and one for the horizontal filter direction.

Note: The syntax element interpolation_filter from the uncompressed header can specify the type of filter to be used for the whole frame. If it is set to SWITCHABLE then the interp_filter syntax element is read from the bitstream for every inter block. {:.alert .alert-info }

RefMvIdx specifies which candidate in the RefStackMv should be used. If there are not enough entries in the stack, a candidate from the RefListMv will be used instead.

drl_mode is a bit sent for candidates in the motion vector stack to indicate if they should be used. drl_mode equal to 0 means to use the current value of idx. drl_mode equal to 1 says to continue searching. DRL stands for "Dynamic Reference List".

Filter Intra Mode Info Semantics

use_filter_intra is a bit specifying whether or not intra filtering can be used.

filter_intra_mode specifies the type of intra filtering, and can take on any of the following values:

filter_intra_mode	Name of filter_intra_mode
0	INTRA_DC_PRED
1	INTRA_V_PRED
2	INTRA_H_PRED
3	INTRA_D153_PRED
4	INTRA_TM_PRED
{:.table .table-sm .table-bordered }

Ref Frames Semantics

comp_mode specifies whether single or compound prediction is used:

comp_mode	Name of comp_mode
0	SINGLE_REFERENCE
1	COMPOUND_REFERENCE
{:.table .table-sm .table-bordered }

SINGLE_REFERENCE indicates that the inter block uses only a single reference frame to generate motion compensated prediction.

COMPOUND_REFERENCE indicates that the inter block uses compound mode.

comp_ref_type is used for compound prediction to specify whether both reference frames come from the same direction or not:

| comp_ref_type | Name of comp_ref_type| Description |:-------------:|:---------------------: ----------- | 0 | UNIDIR_COMP_REFERENCE| Both reference frames from the same direction | 1 | BIDIR_COMP_REFERENCE | One forward and one backward reference frame {:.table .table-sm .table-bordered }

uni_comp_ref, uni_comp_ref_p1, and uni_comp_ref_p2 specify which reference frames are in use when both come from the same direction.

comp_ref, comp_ref_p1, and comp_ref_p2 specify the first reference frame when the two reference frames come from different directions.

comp_bwdref and comp_bwdref_p1 specify the second reference frame when the two reference frames come from different directions.

single_ref_p1, single_ref_p2, single_ref_p3, single_ref_p4, single_ref_p5, and single_ref_p6 specify the reference frame when only a single reference frame is in use.

RefFrame[ 0 ] specifies which frame is used to compute the predicted samples for this block:

RefFrame[ 0 ]	Name of ref_frame
0	INTRA_FRAME
1	LAST_FRAME
2	LAST2_FRAME
3	LAST3_FRAME
4	GOLDEN_FRAME
5	BWDREF_FRAME
6	ALTREF2_FRAME
7	ALTREF_FRAME
{:.table .table-sm .table-bordered }

RefFrame[ 1 ] specifies which additional frame is used in compound prediction:

RefFrame[ 1 ]	Name of ref_frame
-1	NONE (this block uses single prediction)
0	INTRA_FRAME (this block uses interintra prediction)
1	LAST_FRAME
2	LAST2_FRAME
3	LAST3_FRAME
4	GOLDEN_FRAME
5	BWDREF_FRAME
6	ALTREF2_FRAME
7	ALTREF_FRAME
{:.table .table-sm .table-bordered }

Note: Not all combinations of RefFrame[0] and RefFrame[1] can be coded. {:.alert .alert-info }

Read Motion Mode Semantics

use_obmc equal to 1 means that OBMC should be used. use_obmc equal to 0 means that simple translation should be used.

motion_mode specifies the type of motion compensation to perform:

motion_mode	Name of motion_mode
0	SIMPLE
1	OBMC
2	LOCALWARP
{:.table }

Note: A motion_mode equal to SIMPLE is used for blocks requiring global motion. {:.alert .alert-info }

Read Inter Intra Semantics

interintra equal to 1 specifies that an inter prediction should be blended with an intra prediction.

interintra_mode specifies the type of intra prediction to be used:

interintra_mode	Name of interintra_mode
0	II_DC_PRED
1	II_V_PRED
2	II_H_PRED
3	II_SMOOTH_PRED
{:.table .table-sm .table-bordered }

wedge_interintra equal to 1 specifies that wedge blending should be used. wedge_interintra equal to 0 specifies that intra blending should be used.

wedge_index specifies the angle to be used during wedge blending.

Read Compound Type Semantics

comp_group_idx equal to 0 indicates that the compound_idx syntax element should be read. comp_group_idx equal to 1 indicates that the compound_idx syntax element is not present.

compound_idx equal to 1 indicates that a distance based weighted scheme should be used for blending. compound_idx equal to 0 indicates that the averaging scheme should be used for blending.

compound_type specifies how the two predictions should be blended together:

compound_type	Name of compound_type
0	COMPOUND_WEDGE
1	COMPOUND_SEG
2	COMPOUND_AVERAGE
3	COMPOUND_INTRA
4	COMPOUND_DISTANCE
{:.table .table-sm .table-bordered }

Note: COMPOUND_AVERAGE, COMPOUND_INTRA, and COMPOUND_DISTANCE cannot be directly signalled with the compound_type syntax element but are inferred from other syntax elements. {:.alert .alert-info }

wedge_index specifies the angle to be used during wedge blending:

wedge_index	Name of wedge_index
0	WEDGE_HORIZONTAL
1	WEDGE_VERTICAL
2	WEDGE_OBLIQUE27
3	WEDGE_OBLIQUE63
4	WEDGE_OBLIQUE117
5	WEDGE_OBLIQUE153
{:.table .table-sm .table-bordered }

wedge_sign specifies the sign of the wedge blend.

mask_type specifies the type of mask to be used during blending:

mask_type	Name of mask_type
0	UNIFORM_45
1	UNIFORM_45_INV
{:.table .table-sm .table-bordered }

MV Semantics

MvCtx is used to determine which CDFs to use for the motion vector syntax elements.

mv_joint specifies which components of the motion vector difference are non-zero:

mv_joint	Name of mv_joint	Changes row	Changes col
0	MV_JOINT_ZERO	No	No
1	MV_JOINT_HNZVZ	No	Yes
2	MV_JOINT_HZVNZ	Yes	No
3	MV_JOINT_HNZVNZ	Yes	Yes
{:.table .table-sm .table-bordered }

The motion vector difference is added to the PredMv to compute the final motion vector in Mv. It is a requirement of bitstream conformance that the resulting motion vector satisfies -(1 << 14) < Mv[ ref ][ comp ] < (1 << 14) - 1 for comp=0..1.

It is a requirement of bitstream conformance that whenever read_mv returns, the function is_dv_valid would return 1, where is_dv_valid is defined as:

is_dv_valid( ) {
    if ( !use_intrabc ) {
        return 1
    }
    bw = Block_Width[ MiSize ]
    bh = Block_Height[ MiSize ]
    if ( (Mv[ 0 ][ 0 ] & 7) || (Mv[ 0 ][ 1 ] & 7)) {
        return 0
    }
    deltaRow = (Mv[ 0 ][ 0 ] >> 3
    deltaCol = (Mv[ 0 ][ 1 ] >> 3
    srcTopEdge = MiRow + deltaRow
    srcLeftEdge = MiCol + deltaCol
    srcBottomEdge = srcTopEdge + bh
    srcRightEdge = srcLeftEdge + bw
    if ( srcTopEdge < MiRowStart * MI_SIZE ||
         srcLeftEdge < MiColStart * MI_SIZE ||
         srcBottomEdge > MiRowEnd * MI_SIZE ||
         srcRightEdge > MiColEnd * MI_SIZE ) {
        return 0
    }
    sbSize = use_128x128_superblock ? BLOCK_128X128 : BLOCK_64X64
    sbH = Block_Height[ sbSize ]
    activeSbRow = MiRow / sbH
    activeSb64Col = (MiCol * MI_SIZE) >> 6
    srcSbRow = (srcBottomEdge - 1) / sbH
    srcSb64Col = (srcRightEdge - 1) >> 6
    totalSb64PerRow = ((MiColEnd - MiColStart - 1) >> 4) + 1
    activeSb64 = activeSbRow * totalSb64PerRow + activeSb64Col
    srcSb64 = srcSbRow * total_sb64_per_row + srcSb64Col
    if ( srcSb64 >= activeSb64 - INTRABC_DELAY_SB64) {
        return 0
    }
    gradient = 1 + INTRABC_DELAY_SB64 + use_128x128_superblock;
    wfOffset = gradient * (activeSbRow - srcSbRow)
    if (srcSbRow > activeSbRow ||
      srcSb64Col >= activeSb64Col - INTRABC_DELAY_SB64 + wfOffset) {
        return 0
    }
    return 1
}

Note: The purpose of this function is such that, if use_intrabc is equal to 1, the motion vector is additionally constrained in order that the data is fetched from parts of the tile that have already been decoded, and that are not too close to the current block (in order to make a pipelined decoder implementation feasible). {:.alert .alert-info }

MV Component Semantics

mv_sign equal to 0 means that the motion vector difference is positive; mv_sign equal to 1 means that the motion vector difference is negative.

mv_class specifies the class of the motion vector difference. A higher class means that the motion vector difference represents a larger update:

mv_class	Name of mv_class
0	MV_CLASS_0
1	MV_CLASS_1
2	MV_CLASS_2
3	MV_CLASS_3
4	MV_CLASS_4
5	MV_CLASS_5
6	MV_CLASS_6
7	MV_CLASS_7
8	MV_CLASS_8
9	MV_CLASS_9
10	MV_CLASS_10
{:.table .table-sm .table-bordered }

mv_class0_bit specifies the integer part of the motion vector difference. This is only present in the bitstream for class 0 motion vector differences.

mv_class0_fr specifies the first 2 fractional bits of the motion vector difference. This is only present in the bitstream for class 0 motion vector differences.

mv_class0_hp specifies the third fraction bit of the motion vector difference. This is only present in the bitstream for class 0 motion vector differences.

mv_bit specifies bit i of the integer part of the motion vector difference.

mv_fr specifies the first 2 fractional bits of the motion vector difference.

mv_hp specifies the third fractional bit of the motion vector difference.

Residual Semantics

predict_intra is a function call that indicates the conceptual point where intra prediction happens. When this function is called, the intra prediction process specified in [section 7.8.1][] is invoked.

predict_inter is a function call that indicates the conceptual point where inter prediction happens. When this function is called, the inter prediction process specified in [section 7.8.2][] is invoked.

predW and predH are variables containing the smallest size that can be used for inter prediction. (This size may be increased for chroma blocks if not all blocks use inter prediction.)

someUseIntra is a variable that indicates if some of the blocks corresponding to this residual require intra prediction.

Note: The chroma residual block size is always at least 4 in width and height. This means that no transform width or height smaller than 4 is required. As such, a chroma residual may actually cover several luma blocks. Normally, a single prediction is performed for the entire chroma residual block based on the mode info of the bottom right luma block. However, if all the constituent blocks are inter blocks, a special case is triggered and inter prediction is done using a smaller block size for each of the corresponding luma blocks. {:.alert .alert-info }

predict_palette is a function call that indicates the conceptual point where palette prediction happens. When this function is called, the palette prediction process specified in [section 7.8.3][] is invoked.

Note: The predict_inter, predict_intra, predict_palette, and predict_chroma_from_luma functions do not affect the syntax decode process. {:.alert .alert-info }

Transform Block Semantics

reconstruct is a function call that indicates the conceptual point where inverse transform and reconstruction happens. When this function is called, the reconstruction process specified in [section 7.9.2][] is invoked.

predict_chroma_from_luma is a function call that indicates the conceptual point where predicting chroma from luma happens. When this function is called, the predict chroma from luma process specified in [section 7.8.4][] is invoked.

MaxLumaW and MaxLumaH are needed for chroma from luma prediction and store the extent of luma pixels that can be used for prediction.

LoopfilterTxSizes is an array that stores the transform size for each plane and position for use in loop filtering. LoopfilterTxSizes[ plane ][ row ][ col ] stores the transform size where row and col are in units of 4x4 samples.

Note: The transform size is always equal for planes 1 and 2. {:.alert .alert-info }

Coefficients Semantics

TxTypes is an array which stores at a 4x4 pixel granularity the transform type to be used.

Note: The transform type is only read for luma transform blocks, the chroma uses the transform type for the corresponding luma block. {:.alert .alert-info }

Quant is an array storing the quantised coefficients for the current transform block.

all_zero equal to 1 specifies that all coefficients are zero.

eob_extra and eob_extra_bit specify the exact position of the last non-zero coefficient.

Note: The value of eob indicates the index of the end of block. This will be immediately following the last non-zero coefficient. {:.alert .alert-info }

coeff_base_eob is used to compute the base level of the last non-zero coefficient.

coeff_base specifies the base level of a coefficient.

dc_sign specifies the sign of the DC coefficient.

sign_bit specifies the sign of a non-zero coefficient.

coeff_br specifies an increment to the coefficient.

Note: Each quantized coefficient can use coeff_br to provide up to 4 increments. If an increment equal to less than 3 is coded, it signifies that this was the final increment. {:.alert .alert-info }

golomb_length_bit is used to compute the number of extra bits required to code the coefficient.

golomb_data_bit specifies the value of one of the extra bits.

AboveLevelContext and LeftLevelContext are arrays that store at a 4 sample granularity the cumulative sum of coefficient levels.

AboveDcContext and LeftDcContext are arrays that store at a 4 sample granularity 2 bits signalling the sign of the DC coefficient (zero being counted as a separate sign).

Intra Angle Info Semantics

angle_delta_y specifies the offset to be applied to the intra prediction angle specified by the prediction mode in the luma plane, biased by MAX_ANGLE_DELTA so as to encode a positive value.

angle_delta_uv specifies the offset to be applied to the intra prediction angle specified by the prediction mode in the chroma plane biased by MAX_ANGLE_DELTA so as to encode a positive value.

AngleDeltaY is computed from angle_delta_y by removing the MAX_ANGLE_DELTA offset to produce the final luma angle offset value, which may be positive or negative.

AngleDeltaUV is computed from angle_delta_uv by removing the MAX_ANGLE_DELTA offset to produce the final chroma angle offset value, which may be positive or negative.

Read CFL Alphas Semantics

cfl_alpha_signs contains the sign of the alpha values for U and V packed together into a single syntax element with 8 possible values. (The combination of two zero signs is unnecessary.)

signU contains the sign of the alpha value for the U component:

signU	Name of signU
0	CFL_SIGN_ZERO
1	CFL_SIGN_NEG
2	CFL_SIGN_POS
{:.table .table-sm .table-bordered }

signV contains the sign of the alpha value for the V component with the same interpretation as for signU.

cfl_alpha_u contains the absolute value of alpha minus one for the U component.

cfl_alpha_v contains the absolute value of alpha minus one for the V component.

CflAlphaU contains the signed value of the alpha component for the U component.

CflAlphaV contains the signed value of the alpha component for the V component.

Palette Mode Info Semantics

has_palette_y is a boolean value specifying whether a palette is encoded for the Y plane.

has_palette_uv is a boolean value specifying whether a palette is encoded for the UV plane.

palette_size_y_minus_2 is used to compute PaletteSizeY.

PaletteSizeY is a variable holding the Y plane palette size.

palette_size_uv_minus_2 is used to compute PaletteSizeUV.

PaletteSizeUV is a variable holding the UV plane palette size.

use_palette_color_cache_y, if equal to 1, indicates that for a particular palette entry in the luma palette, the cached entry should be used.

use_palette_color_cache_u, if equal to 1, indicates that for a particular palette entry in the U chroma palette, the cached entry should be used.

palette_colors_y is an array holding the Y plane palette colors.

palette_colors_u is an array holding the U plane palette colors.

palette_colors_v is an array holding the V plane palette colors.

delta_encode_palette_colors_v, if equal to 1, indicates that the V chroma palette is encoded using delta encoding.

palette_num_extra_bits_y is used to calculate the number of bits used to store each palette delta value for the luma palette.

palette_num_extra_bits_u is used to calculate the number of bits used to store each palette delta value for the U chroma palette.

palette_num_extra_bits_v is used to calculate the number of bits used to store each palette delta value for the V chroma palette.

palette_delta_y is a delta value for the luma palette.

palette_delta_u is a delta value for the U chroma palette.

palette_delta_v is a delta value for the V chroma palette.

Note: Luma and U delta values give a positive offset relative to the previous palette entry in the same plane. V delta values give a signed offset relative to the U palette entries. {:.alert .alert-info }

palette_delta_sign_bit_v, if equal to 1, indicates that the decoded V chroma palette delta value should be negated.

Palette Tokens Semantics

color_index_map_y holds the index in palette_colors_y for the block's Y plane top left pixel.

color_index_map_uv holds the index in palette_colors_u and palette_colors_v for the block's UV plane top left pixel.

palette_color_idx_y holds the index in ColorOrder for a pixel in the block's Y plane.

palette_color_idx_uv holds the index in ColorOrder for a pixel in the block's UV plane.

Palette Color Context Semantics

ColorOrder is an array holding the mapping from an encoded index to the palette. ColorOrder is ranked in order of frequency of occurrence of each color in the neighbourhood of the current block, weighted by closeness to the current block.

ColorContextHash is a variable derived from the distribution of colors in the neighbourhood of the current block, which is used to determine the probability context used to decode palette_color_idx_y and palette_color_idx_uv.

Read CDEF Semantics

cdef_idx specifies which CDEF filtering parameters should be used for a particular 64 by 64 pixel block. A value of -1 means that CDEF is disabled for that block.

Decode Loop Restoration Unit Semantics

use_wiener specifies if the Wiener filter should be used.

use_sgrproj specifies if the self guided filter should be used.

restoration_type specifies the restoration filter that should be used with the same interpretation as FrameRestorationType.

lr_sgr_set specifies which set of parameters to use for the self guided filter.

subexp_more_bools equal to 0 specifies that the parameter is in the range mk to mk+a-1. subexp_more_bools equal to 0 specifies that the parameter is greater than mk+a-1.

subexp_unif_bools specifies the value of the parameter minus mk.

subexp_bools specifies the value of the parameter minus mk.

Files

07.bitstream.semantics.md

Latest commit

History