HEVC VUI and HRD parameters

Video usability information (VUI, Annex E of H.265) sits at the end of the SPS. It is not needed to decode the pictures, but it tells a player how to show them: shape of the samples, colours, frame rate. The hypothetical reference decoder (HRD) parameters inside it, and in the VPS, describe the buffering model.

VUI fields

Name	Coding	Condition	Description
`aspect_ratio_info_present_flag`	`u(1)`		1: the aspect ratio follows
`aspect_ratio_idc`	`u(8)`	flag above is 1	see sample aspect ratios
`sar_width`, `sar_height`	`u(16)` × 2	`aspect_ratio_idc` is 255	the sample aspect ratio; relatively prime, or 0 for unspecified
`overscan_info_present_flag`	`u(1)`		1: the next flag follows
`overscan_appropriate_flag`	`u(1)`	flag above is 1	1: the picture may be shown with overscan; 0: important content reaches the edges
`video_signal_type_present_flag`	`u(1)`		1: the next three fields follow
`video_format`	`u(3)`	signal type	see video format; 5 when absent
`video_full_range_flag`	`u(1)`	signal type	1: full range (0 to 255 for 8 bits); 0: limited range (16 to 235 luma). 0 when absent
`colour_description_present_flag`	`u(1)`	signal type	1: the three colour codes follow
`colour_primaries`	`u(8)`	colour description	see colour codes; 2 when absent
`transfer_characteristics`	`u(8)`	colour description	2 when absent
`matrix_coeffs`	`u(8)`	colour description	2 when absent
`chroma_loc_info_present_flag`	`u(1)`		1: the chroma sample locations follow
`chroma_sample_loc_type_top_field`	`ue(v)`	flag above is 1	0 to 5, see chroma location
`chroma_sample_loc_type_bottom_field`	`ue(v)`	flag above is 1	0 to 5
`neutral_chroma_indication_flag`	`u(1)`		1: every chroma sample is the mid value, 1 << (bit depth − 1); a grey picture
`field_seq_flag`	`u(1)`		1: each picture is a field, and a picture timing SEI message is in every access unit
`frame_field_info_present_flag`	`u(1)`		1: picture timing SEI messages carry `pic_struct`, `source_scan_type` and `duplicate_flag`
`default_display_window_flag`	`u(1)`		1: the next four offsets follow
`def_disp_win_left_offset`, `…right…`, `…top…`, `…bottom…`	`ue(v)` × 4	flag above is 1	a suggested display area inside the conformance window, in chroma sample units, added to the conformance offsets
`vui_timing_info_present_flag`	`u(1)`		1: the timing fields follow
`vui_num_units_in_tick`	`u(32)`	timing	clock units per tick; equal to the VPS value when that is present
`vui_time_scale`	`u(32)`	timing	clock units per second
`vui_poc_proportional_to_timing_flag`	`u(1)`	timing	1: picture order count is proportional to output time
`vui_num_ticks_poc_diff_one_minus1`	`ue(v)`	flag above is 1	ticks per picture order count step, minus 1
`vui_hrd_parameters_present_flag`	`u(1)`	timing	1: HRD parameters follow, with the common part present
`bitstream_restriction_flag`	`u(1)`		1: the restriction fields follow
`tiles_fixed_structure_flag`	`u(1)`	restrictions	1: every PPS in the sequence has the same tile layout
`motion_vectors_over_pic_boundaries_flag`	`u(1)`	restrictions	0: inter prediction never uses samples outside the picture
`restricted_ref_pic_lists_flag`	`u(1)`	restrictions	1: all P and B slices of a picture have the same list 0, and the same list 1
`min_spatial_segmentation_idc`	`ue(v)`	restrictions	0 to 4095; a bound on the size of slices, tiles or wavefront rows, for parallel decoding
`max_bytes_per_pic_denom`	`ue(v)`	restrictions	0 to 16; 0 means no limit on coded bytes per picture
`max_bits_per_min_cu_denom`	`ue(v)`	restrictions	0 to 16; 0 means no limit on coded bits per coding unit
`log2_max_mv_length_horizontal`	`ue(v)`	restrictions	0 to 15: horizontal motion vectors lie in −2ⁿ to 2ⁿ − 1 quarter samples
`log2_max_mv_length_vertical`	`ue(v)`	restrictions	the same vertically

Picture rate is time_scale ÷ num_units_in_tick pictures per second: unlike H.264, one tick is one picture. 25 frames per second is, for example, 27 000 000 and 1 080 000; 29.97 is 30 000 and 1 001. When field_seq_flag is 1 the pictures are fields, so the rate is a field rate.

Sample aspect ratios

`aspect_ratio_idc`	Sample aspect ratio	Typical use
0	unspecified
1	1:1	square samples: 1920 × 1080, 3840 × 2160, 7680 × 4320
2	12:11	720 × 576 4:3 with overscan
3	10:11	720 × 480 4:3 with overscan
4	16:11	720 × 576 16:9 with overscan
5	40:33	720 × 480 16:9 with overscan
6	24:11	352 × 576 4:3, 480 × 576 16:9 with overscan
7	20:11	352 × 480 4:3, 480 × 480 16:9 with overscan
8	32:11	352 × 576 16:9
9	80:33	352 × 480 16:9
10	18:11	480 × 576 4:3 with overscan
11	15:11	480 × 480 4:3 with overscan
12	64:33	528 × 576 16:9
13	160:99	528 × 480 16:9
14	4:3	1440 × 1080 16:9
15	3:2	1280 × 1080 16:9
16	2:1	960 × 1080 16:9
17–254	reserved; read as 0
255	`EXTENDED_SAR`: `sar_width` : `sar_height`

Video format

`video_format`	Source
0	component
1	PAL
2	NTSC
3	SECAM
4	MAC
5	unspecified
6, 7	reserved; read as 5

Colour codes

Reserved values are read as 2, unspecified.

`colour_primaries`	Primaries
0	reserved
1	BT.709, sRGB: red 0.640, 0.330; green 0.300, 0.600; blue 0.150, 0.060; white D65
2	unspecified
3	reserved
4	BT.470 System M (1953 NTSC, FCC); white C
5	BT.470 System B, G; BT.601 625-line (PAL, SECAM)
6	BT.601 525-line, SMPTE ST 170 (functionally the same as 7)
7	SMPTE ST 240
8	generic film (colour filters, illuminant C)
9	BT.2020, BT.2100: red 0.708, 0.292; green 0.170, 0.797; blue 0.131, 0.046; white D65
10	SMPTE ST 428-1, CIE 1931 XYZ
11	SMPTE RP 431-2, DCI-P3 (white 0.314, 0.351)
12	SMPTE EG 432-1, Display P3 (P3 primaries, white D65)
13–21	reserved
22	EBU Tech. 3213-E
23–255	reserved

`transfer_characteristics`	Transfer
0	reserved
1	BT.709 (functionally the same as 6, 14 and 15)
2	unspecified
3	reserved
4	assumed display gamma 2.2 (BT.470 System M)
5	assumed display gamma 2.8 (BT.470 System B, G)
6	BT.601, SMPTE ST 170
7	SMPTE ST 240
8	linear
9	logarithmic, 100:1 range
10	logarithmic, 100 × √10 : 1 range
11	IEC 61966-2-4 (xvYCC)
12	BT.1361 extended colour gamut
13	IEC 61966-2-1, sRGB and sYCC
14	BT.2020, 10-bit
15	BT.2020, 12-bit
16	SMPTE ST 2084, BT.2100 PQ (HDR10)
17	SMPTE ST 428-1
18	ARIB STD-B67, BT.2100 HLG
19–255	reserved

`matrix_coeffs`	Matrix
0	identity: GBR, as in sRGB and ST 428-1
1	BT.709: K_R 0.2126, K_B 0.0722
2	unspecified
3	reserved
4	FCC: K_R 0.30, K_B 0.11
5	BT.470 B, G; BT.601 625: K_R 0.299, K_B 0.114
6	BT.601 525, SMPTE ST 170: the same coefficients
7	SMPTE ST 240: K_R 0.212, K_B 0.087
8	YCgCo, or YCgCo-R when chroma has one more bit than luma
9	BT.2020 and BT.2100 non-constant luminance: K_R 0.2627, K_B 0.0593
10	BT.2020 constant luminance
11	SMPTE ST 2085, Y′D′_ZD′_X
12	chromaticity-derived non-constant luminance
13	chromaticity-derived constant luminance
14	BT.2100 IC_TC_P
15	IPT-C2 (SMPTE’s IPT-PQ-C2)
16	YCgCo-Re
17	YCgCo-Ro
18–255	reserved

Chroma location

For 4:2:0 only, the position of the top-left chroma sample relative to the top-left luma sample, in luma samples. For 4:2:2 and 4:4:4 the offset is always 0, 0.

Type	Horizontal	Vertical	Common name
0	0	0.5	left (MPEG-2 style); the default
1	0.5	0.5	centre (MPEG-1, JPEG style)
2	0	0	top-left; recommended for BT.2020 and BT.2100
3	0.5	0	top
4	0	1	bottom-left
5	0.5	1	bottom

HRD parameters

hrd_parameters(commonInfPresentFlag, maxNumSubLayersMinus1) appears in the VUI and in the VPS.

Name	Coding	Condition	Description
`nal_hrd_parameters_present_flag`	`u(1)`	common part	1: NAL HRD (whole stream) parameters are given
`vcl_hrd_parameters_present_flag`	`u(1)`	common part	1: VCL HRD (slice data only) parameters are given
`sub_pic_hrd_params_present_flag`	`u(1)`	either flag above	1: the buffer can work on decoding units smaller than a picture
`tick_divisor_minus2`	`u(8)`	sub-picture	the sub-picture clock tick is the clock tick ÷ (this + 2)
`du_cpb_removal_delay_increment_length_minus1`	`u(5)`	sub-picture	bit length of decoding unit removal delays, minus 1
`sub_pic_cpb_params_in_pic_timing_sei_flag`	`u(1)`	sub-picture	1: decoding unit delays are in picture timing SEI; 0: in decoding unit information SEI
`dpb_output_delay_du_length_minus1`	`u(5)`	sub-picture	bit length of `pic_dpb_output_du_delay`, minus 1
`bit_rate_scale`	`u(4)`	either flag	scale for bit rates
`cpb_size_scale`	`u(4)`	either flag	scale for buffer sizes
`cpb_size_du_scale`	`u(4)`	sub-picture	scale for decoding unit buffer sizes
`initial_cpb_removal_delay_length_minus1`	`u(5)`	either flag	bit length of the initial delays in the buffering period SEI, minus 1
`au_cpb_removal_delay_length_minus1`	`u(5)`	either flag	bit length of `au_cpb_removal_delay_minus1` in picture timing SEI, minus 1
`dpb_output_delay_length_minus1`	`u(5)`	either flag	bit length of `pic_dpb_output_delay`, minus 1
`fixed_pic_rate_general_flag[i]`	`u(1)`	per sub-layer	1: constant picture spacing; implies the next flag is 1
`fixed_pic_rate_within_cvs_flag[i]`	`u(1)`	flag above is 0	1: constant picture spacing within each coded video sequence
`elemental_duration_in_tc_minus1[i]`	`ue(v)`	fixed rate	picture spacing in clock ticks, minus 1; 0 to 2047
`low_delay_hrd_flag[i]`	`u(1)`	not fixed rate	1: the HRD works in low-delay mode
`cpb_cnt_minus1[i]`	`ue(v)`	low delay flag is 0	number of alternative buffer specifications, minus 1; 0 to 31
`sub_layer_hrd_parameters(i)`		NAL flag, then again for the VCL flag	see below

Each sub_layer_hrd_parameters repeats, for each of the cpb_cnt_minus1 + 1 specifications:

Name	Coding	Condition	Description
`bit_rate_value_minus1[i]`	`ue(v)`		bit rate = (this + 1) × 2^{6 + bit_rate_scale} bits per second
`cpb_size_value_minus1[i]`	`ue(v)`		buffer size = (this + 1) × 2^{4 + cpb_size_scale} bits
`cpb_size_du_value_minus1[i]`	`ue(v)`	sub-picture	(this + 1) × 2^{4 + cpb_size_du_scale} bits
`bit_rate_du_value_minus1[i]`	`ue(v)`	sub-picture	(this + 1) × 2^{6 + bit_rate_scale} bits per second
`cbr_flag[i]`	`u(1)`		1: constant bit rate