mirror of
https://github.com/ultravideo/uvg266.git
synced 2024-11-27 19:24:06 +00:00
Improve transform tree coding.
- Clean up code and comment. - Change terminology to match H.265 specification where possible. - Move transform splitting for depth==0 out of the coding part. It's not possible to do it here anyway because intra reconstruction is different if the transform is split. - Add checking for transform hierarchy depth when coding split flag. - Fixes bug with cu_data.tr_depth being set. The CU struct was being reused for inter coded CUs, which did not initialize the tr_depth.
This commit is contained in:
parent
d5745aa31a
commit
a38a5fd647
|
@ -582,8 +582,8 @@ void encode_seq_parameter_set(encoder_control* encoder)
|
|||
WRITE_UE(encoder->stream, MAX_DEPTH, "log2_diff_max_min_coding_block_size");
|
||||
WRITE_UE(encoder->stream, 0, "log2_min_transform_block_size_minus2"); // 4x4
|
||||
WRITE_UE(encoder->stream, 3, "log2_diff_max_min_transform_block_size"); // 4x4...32x32
|
||||
WRITE_UE(encoder->stream, 2, "max_transform_hierarchy_depth_inter");
|
||||
WRITE_UE(encoder->stream, 2, "max_transform_hierarchy_depth_intra");
|
||||
WRITE_UE(encoder->stream, TR_DEPTH_INTER, "max_transform_hierarchy_depth_inter");
|
||||
WRITE_UE(encoder->stream, TR_DEPTH_INTRA, "max_transform_hierarchy_depth_intra");
|
||||
|
||||
// Use default scaling list
|
||||
WRITE_U(encoder->stream, ENABLE_SCALING_LIST, 1, "scaling_list_enable_flag");
|
||||
|
@ -1673,51 +1673,68 @@ void encode_transform_tree(encoder_control *encoder, int32_t x_pu, int32_t y_pu,
|
|||
// end Residual Coding
|
||||
}
|
||||
|
||||
/**
|
||||
* \param encoder
|
||||
* \param x_pu Prediction units' x coordinate.
|
||||
* \param y_pu Prediction units' y coordinate.
|
||||
* \param depth Depth from LCU.
|
||||
* \param tr_depth Depth from last CU.
|
||||
* \param parent_coeff_u What was signaled at previous level for cbf_cb.
|
||||
* \param parent_coeff_v What was signlaed at previous level for cbf_cr.
|
||||
*/
|
||||
void encode_transform_coeff(encoder_control *encoder, int32_t x_pu,int32_t y_pu,
|
||||
int8_t depth, int8_t tr_depth, uint8_t parent_coeff_u, uint8_t parent_coeff_v)
|
||||
{
|
||||
int32_t x_cu = x_pu / 2;
|
||||
int32_t y_cu = y_pu / 2;
|
||||
cu_info *cur_cu = &encoder->in.cur_pic->cu_array[MAX_DEPTH][x_cu + y_cu * (encoder->in.width_in_lcu << MAX_DEPTH)];
|
||||
int8_t width = LCU_WIDTH>>depth;
|
||||
int8_t width_c = (depth == MAX_DEPTH + 1 ? width : width >> 1);
|
||||
int8_t split = (cur_cu->tr_depth > depth||!depth);
|
||||
int32_t coeff_fourth = ((LCU_WIDTH>>(depth))*(LCU_WIDTH>>(depth)))+1;
|
||||
int8_t width = LCU_WIDTH >> depth;
|
||||
int8_t width_c = (depth == MAX_PU_DEPTH ? width : width >> 1);
|
||||
|
||||
// NxN signifies implicit transform split at the first transform level.
|
||||
// There is a similar implicit split for inter, but it is only used when
|
||||
// transform hierarchy is not in use.
|
||||
int intra_split_flag = (cur_cu->type == CU_INTRA && cur_cu->part_size == SIZE_NxN);
|
||||
|
||||
// The implicit split by intra NxN is not counted towards max_tr_depth.
|
||||
int max_tr_depth = (cur_cu->type == CU_INTRA ? TR_DEPTH_INTRA + intra_split_flag : TR_DEPTH_INTER);
|
||||
|
||||
int8_t split = cur_cu->tr_depth > depth;
|
||||
|
||||
int8_t cb_flag_u = !split ? cur_cu->coeff_u : cur_cu->coeff_top_u[depth];
|
||||
int8_t cb_flag_v = !split ? cur_cu->coeff_v : cur_cu->coeff_top_v[depth];
|
||||
int intra_split_flag = (cur_cu->type == CU_INTRA && cur_cu->part_size == SIZE_NxN);
|
||||
|
||||
if (depth != 0 && depth != MAX_DEPTH + 1 && !intra_split_flag) {
|
||||
// The split_transform_flag is not signaled when:
|
||||
// - transform size is greater than 32 (depth == 0)
|
||||
// - transform size is 4 (depth == MAX_PU_DEPTH)
|
||||
// - transform depth is max
|
||||
// - cu is intra NxN and it's the first split
|
||||
if (depth > 0 &&
|
||||
depth < MAX_PU_DEPTH &&
|
||||
tr_depth < max_tr_depth &&
|
||||
!(intra_split_flag && tr_depth == 0))
|
||||
{
|
||||
cabac.ctx = &g_trans_subdiv_model[5 - ((g_convert_to_bit[LCU_WIDTH] + 2) - depth)];
|
||||
CABAC_BIN(&cabac,split,"TransformSubdivFlag");
|
||||
CABAC_BIN(&cabac, split, "split_transform_flag");
|
||||
}
|
||||
|
||||
// Signal if chroma data is present
|
||||
// Chroma data is also signaled BEFORE transform split
|
||||
// Chroma data is not signaled if it was set to 0 before split
|
||||
if (tr_depth == 0 || depth < MAX_DEPTH + 1) {
|
||||
// Non-zero chroma U Tcoeffs
|
||||
// Chroma cb flags are not signaled when one of the following:
|
||||
// - transform size is 4 (2x2 chroma transform doesn't exist)
|
||||
// - they have already been signaled to 0 previously
|
||||
// When they are not present they are inferred to be 0, except for size 4
|
||||
// when the flags from previous level are used.
|
||||
if (depth < MAX_PU_DEPTH) {
|
||||
cabac.ctx = &g_qt_cbf_model_chroma[tr_depth];
|
||||
|
||||
if (tr_depth == 0 || parent_coeff_u) {
|
||||
CABAC_BIN(&cabac, cb_flag_u, "cbf_chroma_u");
|
||||
if (tr_depth == 0 || parent_coeff_u) {
|
||||
CABAC_BIN(&cabac, cb_flag_u, "cbf_cb");
|
||||
}
|
||||
|
||||
// Non-zero chroma V Tcoeffs
|
||||
// NOTE: Using the same ctx as before
|
||||
|
||||
if (tr_depth == 0 || parent_coeff_v) {
|
||||
CABAC_BIN(&cabac, cb_flag_v, "cbf_chroma_v");
|
||||
if (tr_depth == 0 || parent_coeff_v) {
|
||||
CABAC_BIN(&cabac, cb_flag_v, "cbf_cr");
|
||||
}
|
||||
}
|
||||
|
||||
if (split) {
|
||||
uint8_t offset = 1<<(MAX_DEPTH-1-depth);
|
||||
uint8_t pu_offset = 1<<(MAX_PU_DEPTH-1-depth);
|
||||
cu_info *cu_a = &encoder->in.cur_pic->cu_array[MAX_DEPTH][x_cu + offset + y_cu * (encoder->in.width_in_lcu << MAX_DEPTH)];
|
||||
cu_info *cu_b = &encoder->in.cur_pic->cu_array[MAX_DEPTH][x_cu + (y_cu + offset) * (encoder->in.width_in_lcu << MAX_DEPTH)];
|
||||
cu_info *cu_c = &encoder->in.cur_pic->cu_array[MAX_DEPTH][x_cu + offset + (y_cu + offset) * (encoder->in.width_in_lcu << MAX_DEPTH)];
|
||||
uint8_t pu_offset = 1 << (MAX_PU_DEPTH - (depth + 1));
|
||||
encode_transform_coeff(encoder, x_pu, y_pu, depth + 1, tr_depth + 1, cb_flag_u, cb_flag_v);
|
||||
encode_transform_coeff(encoder, x_pu + pu_offset, y_pu, depth + 1, tr_depth + 1, cb_flag_u, cb_flag_v);
|
||||
encode_transform_coeff(encoder, x_pu, y_pu + pu_offset, depth + 1, tr_depth + 1, cb_flag_u, cb_flag_v);
|
||||
|
@ -1725,12 +1742,18 @@ void encode_transform_coeff(encoder_control *encoder, int32_t x_pu,int32_t y_pu,
|
|||
return;
|
||||
}
|
||||
|
||||
if(cur_cu->type == CU_INTRA || tr_depth || cur_cu->coeff_u || cur_cu->coeff_v) {
|
||||
// Non-zero luma Tcoeffs
|
||||
// Luma coded block flag is signaled when one of the following:
|
||||
// - prediction mode is intra
|
||||
// - transform depth > 0
|
||||
// - we have chroma coefficients at this level
|
||||
// When it is not present, it is inferred to be 1.
|
||||
if(cur_cu->type == CU_INTRA || tr_depth > 0 || cur_cu->coeff_u || cur_cu->coeff_v) {
|
||||
cabac.ctx = &g_qt_cbf_model_luma[!tr_depth];
|
||||
CABAC_BIN(&cabac, cur_cu->coeff_y, "cbf_luma");
|
||||
}
|
||||
|
||||
// End of transform tree.
|
||||
// Beginning of transform unit.
|
||||
|
||||
{
|
||||
coefficient coeff_y[LCU_WIDTH*LCU_WIDTH+1];
|
||||
|
|
|
@ -56,6 +56,9 @@ typedef int16_t coefficient;
|
|||
#define MIN_SIZE 3 /*!< log2_min_coding_block_size */
|
||||
#define CU_MIN_SIZE_PIXELS 8 /*!< pow(2, MIN_SIZE) */
|
||||
|
||||
#define TR_DEPTH_INTRA 2
|
||||
#define TR_DEPTH_INTER 2
|
||||
|
||||
#define ENABLE_PCM 0 /*!< Setting to 1 will enable using PCM blocks (current intra-search does not consider PCM) */
|
||||
#define ENABLE_SIGN_HIDING 1
|
||||
#define ENABLE_SCALING_LIST 0 /*!< Enable usage of (default) scaling list */
|
||||
|
|
|
@ -33,6 +33,7 @@ void inter_set_block(picture* pic, uint32_t x_cu, uint32_t y_cu, uint8_t depth,
|
|||
// Width in smallest CU
|
||||
int width_in_scu = pic->width_in_lcu<<MAX_DEPTH;
|
||||
int block_scu_width = (LCU_WIDTH>>depth)/(LCU_WIDTH>>MAX_DEPTH);
|
||||
int tr_depth = (depth == 0 ? 1 : depth);
|
||||
// Loop through all the block in the area of cur_cu
|
||||
for (y = y_cu; y < y_cu + block_scu_width; y++) {
|
||||
int cu_pos = y * width_in_scu; //!< calculate y-position once, use with every x
|
||||
|
@ -44,6 +45,7 @@ void inter_set_block(picture* pic, uint32_t x_cu, uint32_t y_cu, uint8_t depth,
|
|||
pic->cu_array[MAX_DEPTH][cu_pos + x].inter.mv[0] = cur_cu->inter.mv[0];
|
||||
pic->cu_array[MAX_DEPTH][cu_pos + x].inter.mv[1] = cur_cu->inter.mv[1];
|
||||
pic->cu_array[MAX_DEPTH][cu_pos + x].inter.mv_dir = cur_cu->inter.mv_dir;
|
||||
pic->cu_array[MAX_DEPTH][cu_pos + x].tr_depth = tr_depth;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
@ -38,6 +38,7 @@ void intra_set_block_mode(picture *pic,uint32_t x_cu, uint32_t y_cu, uint8_t dep
|
|||
uint32_t x, y;
|
||||
int width_in_scu = pic->width_in_lcu<<MAX_DEPTH; //!< Width in smallest CU
|
||||
int block_scu_width = (LCU_WIDTH>>depth)/(LCU_WIDTH>>MAX_DEPTH);
|
||||
int tr_depth = (part_mode == SIZE_2Nx2N ? depth : depth + 1);
|
||||
|
||||
// Loop through all the blocks in the area of cur_cu
|
||||
for (y = y_cu; y < y_cu + block_scu_width; y++) {
|
||||
|
@ -47,7 +48,7 @@ void intra_set_block_mode(picture *pic,uint32_t x_cu, uint32_t y_cu, uint8_t dep
|
|||
pic->cu_array[MAX_DEPTH][cu_pos + x].type = CU_INTRA;
|
||||
pic->cu_array[MAX_DEPTH][cu_pos + x].intra[0].mode = mode;
|
||||
pic->cu_array[MAX_DEPTH][cu_pos + x].part_size = part_mode;
|
||||
//pic->cu_array[MAX_DEPTH][cu_pos + x].tr_depth = depth + 1;
|
||||
pic->cu_array[MAX_DEPTH][cu_pos + x].tr_depth = tr_depth;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
Loading…
Reference in a new issue