generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Calculate rd cost per transformed block

Browse Source 
This commit makes the rate-distortion optimization loop evaluate
the rd costs of regular quantization and all zero coeffs, per
transformed block. It improves speed 1 compression performance:

derf: 0.245%
yt:   0.515%

For a large partition that consists multiple transformed blocks,
this allows more flexibility to selectively force a portion of
them coded as all zero coeffs, as well be continued in the next
patches.

Change-Id: I211518be4179747b57375696f017d1160cc91851
This commit is contained in:
Jingning Han
2013-09-20 11:30:46 -07:00
parent 24df77e951
commit 78fbb10642
1 changed files with 40 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

60
vp9/encoder/vp9_rdopt.c
View File

@@ -554,9 +554,13 @@ struct rdcost_block_args {
TX_SIZE tx_size; TX_SIZE tx_size;
int bw; int bw;
int bh; int bh;
int rate; int rate[256];
int64_t dist; int64_t dist[256];
int64_t sse; int64_t sse[256];
int this_rate;
int64_t this_dist;
int64_t this_sse;
int64_t this_rd;
int64_t best_rd; int64_t best_rd;
int skip; int skip;
const int16_t *scan, *nb; const int16_t *scan, *nb;
@@ -573,17 +577,17 @@ static void dist_block(int plane, int block, TX_SIZE tx_size, void *arg) {
int shift = args->tx_size == TX_32X32 ? 0 : 2; int shift = args->tx_size == TX_32X32 ? 0 : 2;
int16_t *const coeff = BLOCK_OFFSET(p->coeff, block); int16_t *const coeff = BLOCK_OFFSET(p->coeff, block);
int16_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block); int16_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
args->dist += vp9_block_error(coeff, dqcoeff, 16 << ss_txfrm_size, args->dist[block] = vp9_block_error(coeff, dqcoeff, 16 << ss_txfrm_size,
&this_sse) >> shift; &this_sse) >> shift;
args->sse += this_sse >> shift; args->sse[block] = this_sse >> shift;
if (x->skip_encode && if (x->skip_encode &&
xd->this_mi->mbmi.ref_frame[0] == INTRA_FRAME) { xd->this_mi->mbmi.ref_frame[0] == INTRA_FRAME) {
// TODO(jingning): tune the model to better capture the distortion. // TODO(jingning): tune the model to better capture the distortion.
int64_t p = (pd->dequant[1] * pd->dequant[1] * int64_t p = (pd->dequant[1] * pd->dequant[1] *
(1 << ss_txfrm_size)) >> shift; (1 << ss_txfrm_size)) >> shift;
args->dist += p; args->dist[block] = p;
args->sse += p; args->sse[block] = p;
} }
} }
@@ -594,7 +598,7 @@ static void rate_block(int plane, int block, BLOCK_SIZE plane_bsize,
int x_idx, y_idx; int x_idx, y_idx;
txfrm_block_to_raster_xy(plane_bsize, args->tx_size, block, &x_idx, &y_idx); txfrm_block_to_raster_xy(plane_bsize, args->tx_size, block, &x_idx, &y_idx);
args->rate += cost_coeffs(args->x, plane, block, args->rate[block] = cost_coeffs(args->x, plane, block,
args->t_above + x_idx, args->t_above + x_idx,
args->t_left + y_idx, args->tx_size, args->t_left + y_idx, args->tx_size,
args->scan, args->nb); args->scan, args->nb);
@@ -610,16 +614,6 @@ static void block_yrd_txfm(int plane, int block, BLOCK_SIZE plane_bsize,
if (args->skip) if (args->skip)
return; return;
rd1 = RDCOST(x->rdmult, x->rddiv, args->rate, args->dist);
rd2 = RDCOST(x->rdmult, x->rddiv, 0, args->sse);
rd = MIN(rd1, rd2);
if (rd > args->best_rd) {
args->skip = 1;
args->rate = INT_MAX;
args->dist = INT64_MAX;
args->sse = INT64_MAX;
return;
}
if (!is_inter_block(&xd->this_mi->mbmi)) if (!is_inter_block(&xd->this_mi->mbmi))
vp9_encode_block_intra(plane, block, plane_bsize, tx_size, &encode_args); vp9_encode_block_intra(plane, block, plane_bsize, tx_size, &encode_args);
@@ -628,6 +622,18 @@ static void block_yrd_txfm(int plane, int block, BLOCK_SIZE plane_bsize,
dist_block(plane, block, tx_size, args); dist_block(plane, block, tx_size, args);
rate_block(plane, block, plane_bsize, tx_size, args); rate_block(plane, block, plane_bsize, tx_size, args);
rd1 = RDCOST(x->rdmult, x->rddiv, args->rate[block], args->dist[block]);
rd2 = RDCOST(x->rdmult, x->rddiv, 0, args->sse[block]);
rd = MIN(rd1, rd2);
args->this_rate += args->rate[block];
args->this_dist += args->dist[block];
args->this_sse += args->sse[block];
args->this_rd += rd;
if (args->this_rd > args->best_rd) {
args->skip = 1;
return;
}
} }
static void txfm_rd_in_plane(MACROBLOCK *x, static void txfm_rd_in_plane(MACROBLOCK *x,
@@ -643,7 +649,8 @@ static void txfm_rd_in_plane(MACROBLOCK *x,
int i; int i;
struct rdcost_block_args args = { x, { 0 }, { 0 }, tx_size, struct rdcost_block_args args = { x, { 0 }, { 0 }, tx_size,
num_4x4_blocks_wide, num_4x4_blocks_high, num_4x4_blocks_wide, num_4x4_blocks_high,
0, 0, 0, ref_best_rd, 0 }; { 0 }, { 0 }, { 0 },
0, 0, 0, 0, ref_best_rd, 0 };
if (plane == 0) if (plane == 0)
xd->this_mi->mbmi.tx_size = tx_size; xd->this_mi->mbmi.tx_size = tx_size;
@@ -685,10 +692,17 @@ static void txfm_rd_in_plane(MACROBLOCK *x,
} }
foreach_transformed_block_in_plane(xd, bsize, plane, block_yrd_txfm, &args); foreach_transformed_block_in_plane(xd, bsize, plane, block_yrd_txfm, &args);
*distortion = args.dist; if (args.skip) {
*rate = args.rate; *rate = INT_MAX;
*sse = args.sse; *distortion = INT64_MAX;
*skippable = vp9_is_skippable_in_plane(xd, bsize, plane) && (!args.skip); *sse = INT64_MAX;
*skippable = 0;
} else {
*distortion = args.this_dist;
*rate = args.this_rate;
*sse = args.this_sse;
*skippable = vp9_is_skippable_in_plane(xd, bsize, plane);
}
} }
static void choose_largest_txfm_size(VP9_COMP *cpi, MACROBLOCK *x, static void choose_largest_txfm_size(VP9_COMP *cpi, MACROBLOCK *x,