[FFmpeg-devel] [PATCH 1/2] avformat/hevc: add support for writing alpha layer
Timo Rothenpieler
timo at rothenpieler.org
Tue Dec 10 21:16:31 EET 2024
---
libavformat/hevc.c | 138 ++++++++++++++++++++++++++++++++++++++++-----
1 file changed, 123 insertions(+), 15 deletions(-)
diff --git a/libavformat/hevc.c b/libavformat/hevc.c
index 7cf0b0ffb2..50e7f91bd2 100644
--- a/libavformat/hevc.c
+++ b/libavformat/hevc.c
@@ -82,6 +82,8 @@ typedef struct HEVCDecoderConfigurationRecord {
uint8_t lengthSizeMinusOne;
uint8_t numOfArrays;
HVCCNALUnitArray arrays[NB_ARRAYS];
+
+ uint8_t alpha_layer_nuh_id;
} HEVCDecoderConfigurationRecord;
typedef struct HVCCProfileTierLevel {
@@ -149,6 +151,7 @@ static void hvcc_update_ptl(HEVCDecoderConfigurationRecord *hvcc,
static void hvcc_parse_ptl(GetBitContext *gb,
HEVCDecoderConfigurationRecord *hvcc,
+ int profile_present_flag,
unsigned int max_sub_layers_minus1)
{
unsigned int i;
@@ -156,13 +159,17 @@ static void hvcc_parse_ptl(GetBitContext *gb,
uint8_t sub_layer_profile_present_flag[HEVC_MAX_SUB_LAYERS];
uint8_t sub_layer_level_present_flag[HEVC_MAX_SUB_LAYERS];
- general_ptl.profile_space = get_bits(gb, 2);
- general_ptl.tier_flag = get_bits1(gb);
- general_ptl.profile_idc = get_bits(gb, 5);
- general_ptl.profile_compatibility_flags = get_bits_long(gb, 32);
- general_ptl.constraint_indicator_flags = get_bits64(gb, 48);
- general_ptl.level_idc = get_bits(gb, 8);
- hvcc_update_ptl(hvcc, &general_ptl);
+ if (profile_present_flag) {
+ general_ptl.profile_space = get_bits(gb, 2);
+ general_ptl.tier_flag = get_bits1(gb);
+ general_ptl.profile_idc = get_bits(gb, 5);
+ general_ptl.profile_compatibility_flags = get_bits_long(gb, 32);
+ general_ptl.constraint_indicator_flags = get_bits64(gb, 48);
+ general_ptl.level_idc = get_bits(gb, 8);
+ hvcc_update_ptl(hvcc, &general_ptl);
+ } else {
+ skip_bits(gb, 8); // general_level_idc
+ }
for (i = 0; i < max_sub_layers_minus1; i++) {
sub_layer_profile_present_flag[i] = get_bits1(gb);
@@ -387,12 +394,16 @@ static void skip_sub_layer_ordering_info(GetBitContext *gb)
static int hvcc_parse_vps(GetBitContext *gb, HVCCNALUnit *nal,
HEVCDecoderConfigurationRecord *hvcc)
{
+ uint8_t vps_base_layer_internal_flag, vps_max_layers_minus1;
+ uint8_t vps_sub_layer_ordering_info_present_flag, vps_max_layer_id;
+ int vps_num_layer_sets_minus1;
+ int i, j;
+
nal->parameter_set_id = get_bits(gb, 4);
- /*
- * vps_reserved_three_2bits u(2)
- * vps_max_layers_minus1 u(6)
- */
- skip_bits(gb, 8);
+
+ vps_base_layer_internal_flag = get_bits(gb, 1);
+ skip_bits(gb, 1); // vps_base_layer_available_flag
+ vps_max_layers_minus1 = get_bits(gb, 6);
nal->vps_max_sub_layers_minus1 = get_bits(gb, 3);
@@ -413,7 +424,104 @@ static int hvcc_parse_vps(GetBitContext *gb, HVCCNALUnit *nal,
*/
skip_bits(gb, 17);
- hvcc_parse_ptl(gb, hvcc, nal->vps_max_sub_layers_minus1);
+ hvcc_parse_ptl(gb, hvcc, 1, nal->vps_max_sub_layers_minus1);
+
+ vps_sub_layer_ordering_info_present_flag = get_bits(gb, 1);
+ for (i = (vps_sub_layer_ordering_info_present_flag ? 0 : nal->vps_max_sub_layers_minus1);
+ i <= nal->vps_max_sub_layers_minus1; i++) {
+ get_ue_golomb(gb); // vps_max_dec_pic_buffering_minus1
+ get_ue_golomb(gb); // vps_max_num_reorder_pics
+ get_ue_golomb(gb); // vps_max_latency_increase_plus1
+ }
+
+ vps_max_layer_id = get_bits(gb, 6);
+ vps_num_layer_sets_minus1 = get_ue_golomb(gb);
+ skip_bits_long(gb, (vps_max_layer_id + 1) * vps_num_layer_sets_minus1); // layer_id_included_flag[i][j]
+
+ if (get_bits(gb, 1)) { // vps_timing_info_present_flag
+ int vps_num_hrd_parameters;
+
+ skip_bits_long(gb, 64); // vps_num_units_in_tick, vps_time_scale
+
+ if (get_bits(gb, 1)) // vps_poc_proportional_to_timing_flag
+ get_ue_golomb(gb); // vps_num_ticks_poc_diff_one_minus1
+
+ vps_num_hrd_parameters = get_ue_golomb(gb); // vps_num_hrd_parameters
+
+ for (i = 0; i < vps_num_hrd_parameters; i++) {
+ int cprms_present_flag;
+
+ get_ue_golomb(gb); // hrd_layer_set_idx[i]
+ if (i > 0)
+ cprms_present_flag = get_bits(gb, 1);
+ else
+ cprms_present_flag = 1;
+
+ skip_hrd_parameters(gb, cprms_present_flag, nal->vps_max_sub_layers_minus1);
+ }
+ }
+
+ if (get_bits(gb, 1)) { // vps_extension_flag
+ uint8_t num_scalability_types = 0;
+ uint8_t max_layers_minus_1 = FFMIN(62, vps_max_layers_minus1);
+ uint8_t splitting_flag, vps_nuh_layer_id_present_flag;
+ uint8_t scalability_mask_flag[16] = { 0 };
+ uint8_t dimension_id_len[16] = { 0 };
+ uint8_t layer_id_in_nuh[64] = { 0 };
+
+ align_get_bits(gb);
+
+ if (vps_max_layers_minus1 > 0 && vps_base_layer_internal_flag)
+ hvcc_parse_ptl(gb, hvcc, 0, nal->vps_max_sub_layers_minus1);
+
+ splitting_flag = get_bits(gb, 1);
+
+ for (i = 0; i < 16; i++)
+ if (get_bits(gb, 1))
+ scalability_mask_flag[num_scalability_types++] = i;
+
+ for (j = 0; j < (num_scalability_types - splitting_flag); j++)
+ dimension_id_len[j] = get_bits(gb, 3) + 1;
+
+ vps_nuh_layer_id_present_flag = get_bits(gb, 1);
+
+ for (i = 1; i <= max_layers_minus_1; i++) {
+ if (vps_nuh_layer_id_present_flag)
+ layer_id_in_nuh[i] = get_bits(gb, 6);
+ else
+ layer_id_in_nuh[i] = i;
+
+ if (!splitting_flag) {
+ for (j = 0; j < num_scalability_types; j++) {
+ int dimension_id = get_bits(gb, dimension_id_len[j]);
+
+ if (dimension_id == 1 /* AUX_ALPHA */ && scalability_mask_flag[j] == 3 /* AuxId */)
+ hvcc->alpha_layer_nuh_id = layer_id_in_nuh[i];
+ }
+ }
+ }
+
+ if (splitting_flag) {
+ uint8_t dim_bit_offset[17] = { 0 };
+
+ dim_bit_offset[0] = 0;
+ for (j = 1; j < num_scalability_types; j++)
+ dim_bit_offset[j] = dim_bit_offset[j-1] + dimension_id_len[j-1];
+ dim_bit_offset[num_scalability_types] = 6;
+
+ if (num_scalability_types > 0 && dim_bit_offset[num_scalability_types - 1] >= 6)
+ return 0; // invalid bitstream
+
+ for (i = 1; i <= max_layers_minus_1; i++) {
+ for (j = 0; j < num_scalability_types; j++) {
+ int dimension_id = (layer_id_in_nuh[i] & ((1 << dim_bit_offset[j+1]) - 1)) >> dim_bit_offset[j];
+
+ if (dimension_id == 1 /* AUX_ALPHA */ && scalability_mask_flag[j] == 3 /* AuxId */)
+ hvcc->alpha_layer_nuh_id = layer_id_in_nuh[i];
+ }
+ }
+ }
+ }
/* nothing useful for hvcC past this point */
return 0;
@@ -551,7 +659,7 @@ static int hvcc_parse_sps(GetBitContext *gb, HVCCNALUnit *nal,
if (!multi_layer_ext_sps_flag) {
hvcc->temporalIdNested = get_bits1(gb);
- hvcc_parse_ptl(gb, hvcc, sps_max_sub_layers_minus1);
+ hvcc_parse_ptl(gb, hvcc, 1, sps_max_sub_layers_minus1);
}
nal->parameter_set_id = get_ue_golomb_long(gb);
@@ -762,7 +870,7 @@ static int hvcc_add_nal_unit(const uint8_t *nal_buf, uint32_t nal_size,
goto end;
nal_unit_parse_header(&gbc, &nal_type, &nuh_layer_id);
- if (!is_lhvc && nuh_layer_id > 0)
+ if (!is_lhvc && nuh_layer_id > 0 && nuh_layer_id != hvcc->alpha_layer_nuh_id)
goto end;
/*
--
2.45.2
More information about the ffmpeg-devel
mailing list