Merge "Add a conservative RD based active map in vp9."
diff --git a/examples.mk b/examples.mk
index 5c6e42d..7ec1a8e 100644
--- a/examples.mk
+++ b/examples.mk
@@ -140,13 +140,13 @@
endif
decode_with_partial_drops.GUID = 61C2D026-5754-46AC-916F-1343ECC5537E
decode_with_partial_drops.DESCRIPTION = Drops parts of frames while decoding
-EXAMPLES-$(CONFIG_VP8_ENCODER) += vp8_set_maps.c
-vp8_set_maps.SRCS += ivfenc.h ivfenc.c
-vp8_set_maps.SRCS += tools_common.h tools_common.c
-vp8_set_maps.SRCS += video_common.h
-vp8_set_maps.SRCS += video_writer.h video_writer.c
-vp8_set_maps.GUID = ECB2D24D-98B8-4015-A465-A4AF3DCC145F
-vp8_set_maps.DESCRIPTION = VP8 set active and ROI maps
+EXAMPLES-$(CONFIG_ENCODERS) += set_maps.c
+set_maps.SRCS += ivfenc.h ivfenc.c
+set_maps.SRCS += tools_common.h tools_common.c
+set_maps.SRCS += video_common.h
+set_maps.SRCS += video_writer.h video_writer.c
+set_maps.GUID = ECB2D24D-98B8-4015-A465-A4AF3DCC145F
+set_maps.DESCRIPTION = Set active and ROI maps
EXAMPLES-$(CONFIG_VP8_ENCODER) += vp8cx_set_ref.c
vp8cx_set_ref.SRCS += ivfenc.h ivfenc.c
vp8cx_set_ref.SRCS += tools_common.h tools_common.c
diff --git a/examples/vp8_set_maps.c b/examples/set_maps.c
similarity index 87%
rename from examples/vp8_set_maps.c
rename to examples/set_maps.c
index f3cc9a7..4343832 100644
--- a/examples/vp8_set_maps.c
+++ b/examples/set_maps.c
@@ -56,7 +56,8 @@
static const char *exec_name;
void usage_exit() {
- fprintf(stderr, "Usage: %s <width> <height> <infile> <outfile>\n", exec_name);
+ fprintf(stderr, "Usage: %s <codec> <width> <height> <infile> <outfile>\n",
+ exec_name);
exit(EXIT_FAILURE);
}
@@ -65,8 +66,8 @@
unsigned int i;
vpx_roi_map_t roi = {0};
- roi.rows = cfg->g_h / 16;
- roi.cols = cfg->g_w / 16;
+ roi.rows = (cfg->g_h + 15) / 16;
+ roi.cols = (cfg->g_w + 15) / 16;
roi.delta_q[0] = 0;
roi.delta_q[1] = -2;
@@ -98,8 +99,8 @@
unsigned int i;
vpx_active_map_t map = {0};
- map.rows = cfg->g_h / 16;
- map.cols = cfg->g_w / 16;
+ map.rows = (cfg->g_h + 15) / 16;
+ map.cols = (cfg->g_w + 15) / 16;
map.active_map = (uint8_t *)malloc(map.rows * map.cols);
for (i = 0; i < map.rows * map.cols; ++i)
@@ -115,8 +116,8 @@
vpx_codec_ctx_t *codec) {
vpx_active_map_t map = {0};
- map.rows = cfg->g_h / 16;
- map.cols = cfg->g_w / 16;
+ map.rows = (cfg->g_h + 15) / 16;
+ map.cols = (cfg->g_w + 15) / 16;
map.active_map = NULL;
if (vpx_codec_control(codec, VP8E_SET_ACTIVEMAP, &map))
@@ -161,20 +162,20 @@
VpxVideoWriter *writer = NULL;
const VpxInterface *encoder = NULL;
const int fps = 2; // TODO(dkovalev) add command line argument
- const int bitrate = 200; // kbit/s TODO(dkovalev) add command line argument
+ const double bits_per_pixel_per_frame = 0.067;
exec_name = argv[0];
- if (argc != 5)
+ if (argc != 6)
die("Invalid number of arguments");
- encoder = get_vpx_encoder_by_name("vp8"); // only vp8 for now
+ encoder = get_vpx_encoder_by_name(argv[1]);
if (!encoder)
die("Unsupported codec.");
info.codec_fourcc = encoder->fourcc;
- info.frame_width = strtol(argv[1], NULL, 0);
- info.frame_height = strtol(argv[2], NULL, 0);
+ info.frame_width = strtol(argv[2], NULL, 0);
+ info.frame_height = strtol(argv[3], NULL, 0);
info.time_base.numerator = 1;
info.time_base.denominator = fps;
@@ -200,14 +201,16 @@
cfg.g_h = info.frame_height;
cfg.g_timebase.num = info.time_base.numerator;
cfg.g_timebase.den = info.time_base.denominator;
- cfg.rc_target_bitrate = bitrate;
+ cfg.rc_target_bitrate = (unsigned int)(bits_per_pixel_per_frame * cfg.g_w *
+ cfg.g_h * fps / 1000);
+ cfg.g_lag_in_frames = 0;
- writer = vpx_video_writer_open(argv[4], kContainerIVF, &info);
+ writer = vpx_video_writer_open(argv[5], kContainerIVF, &info);
if (!writer)
- die("Failed to open %s for writing.", argv[4]);
+ die("Failed to open %s for writing.", argv[5]);
- if (!(infile = fopen(argv[3], "rb")))
- die("Failed to open %s for reading.", argv[3]);
+ if (!(infile = fopen(argv[4], "rb")))
+ die("Failed to open %s for reading.", argv[4]);
if (vpx_codec_enc_init(&codec, encoder->interface(), &cfg, 0))
die_codec(&codec, "Failed to initialize encoder");
@@ -215,7 +218,7 @@
while (vpx_img_read(&raw, infile)) {
++frame_count;
- if (frame_count == 22) {
+ if (frame_count == 22 && encoder->fourcc == VP8_FOURCC) {
set_roi_map(&cfg, &codec);
} else if (frame_count == 33) {
set_active_map(&cfg, &codec);
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index a52171c..9cc43f9 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -2284,7 +2284,7 @@
if (cpi->sf.tx_size_search_method == USE_LARGESTALL) {
return ALLOW_32X32;
} else if (cpi->sf.tx_size_search_method == USE_FULL_RD) {
- const int frame_type = get_frame_type(cpi);
+ const MV_REFERENCE_FRAME frame_type = get_frame_type(cpi);
return cpi->rd_tx_select_threshes[frame_type][ALLOW_32X32] >
cpi->rd_tx_select_threshes[frame_type][TX_MODE_SELECT] ?
ALLOW_32X32 : TX_MODE_SELECT;
diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c
index f548de5..acb5a09 100644
--- a/vp9/encoder/vp9_firstpass.c
+++ b/vp9/encoder/vp9_firstpass.c
@@ -1412,7 +1412,7 @@
active_max_gf_interval = rc->max_gf_interval;
i = 0;
- while (i < twopass->static_scene_max_gf_interval && i < rc->frames_to_key) {
+ while (i < rc->static_scene_max_gf_interval && i < rc->frames_to_key) {
++i;
// Accumulate error score of frames in this gf group.
diff --git a/vp9/encoder/vp9_firstpass.h b/vp9/encoder/vp9_firstpass.h
index 9f44a30..278b22c 100644
--- a/vp9/encoder/vp9_firstpass.h
+++ b/vp9/encoder/vp9_firstpass.h
@@ -57,7 +57,6 @@
double modified_error_left;
double kf_intra_err_min;
double gf_intra_err_min;
- int static_scene_max_gf_interval;
int kf_bits;
// Remaining error from uncoded frames in a gf group. Two pass use only
int64_t gf_group_error_left;
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index 57d2c78..0a5033f 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -97,7 +97,7 @@
static const double in_frame_q_adj_ratio[MAX_SEGMENTS] =
{1.0, 2.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0};
-static INLINE void Scale2Ratio(int mode, int *hr, int *hs) {
+static INLINE void Scale2Ratio(VPX_SCALING mode, int *hr, int *hs) {
switch (mode) {
case NORMAL:
*hr = 1;
@@ -707,7 +707,6 @@
if (speed >= 5) {
sf->comp_inter_joint_search_thresh = BLOCK_SIZES;
sf->partition_search_type = FIXED_PARTITION;
- sf->always_this_block_size = BLOCK_16X16;
sf->tx_size_search_method = frame_is_intra_only(cm) ?
USE_FULL_RD : USE_LARGESTALL;
sf->mode_search_skip_flags = FLAG_SKIP_INTRA_DIRMISMATCH |
@@ -943,6 +942,9 @@
for (i = 0; i < BLOCK_SIZES; ++i)
sf->disable_inter_mode_mask[i] = 0;
sf->max_intra_bsize = BLOCK_64X64;
+ // This setting only takes effect when partition_search_type is set
+ // to FIXED_PARTITION.
+ sf->always_this_block_size = BLOCK_16X16;
switch (cpi->oxcf.mode) {
case MODE_BESTQUALITY:
@@ -1134,21 +1136,19 @@
void vp9_new_framerate(VP9_COMP *cpi, double framerate) {
VP9_COMMON *const cm = &cpi->common;
+ RATE_CONTROL *const rc = &cpi->rc;
+ VP9_CONFIG *const oxcf = &cpi->oxcf;
int vbr_max_bits;
- if (framerate < 0.1)
- framerate = 30;
-
- cpi->oxcf.framerate = framerate;
+ oxcf->framerate = framerate < 0.1 ? 30 : framerate;
cpi->output_framerate = cpi->oxcf.framerate;
- cpi->rc.av_per_frame_bandwidth = (int)(cpi->oxcf.target_bandwidth
- / cpi->output_framerate);
- cpi->rc.min_frame_bandwidth = (int)(cpi->rc.av_per_frame_bandwidth *
- cpi->oxcf.two_pass_vbrmin_section / 100);
+ rc->av_per_frame_bandwidth = (int)(oxcf->target_bandwidth /
+ cpi->output_framerate);
+ rc->min_frame_bandwidth = (int)(rc->av_per_frame_bandwidth *
+ oxcf->two_pass_vbrmin_section / 100);
- cpi->rc.min_frame_bandwidth = MAX(cpi->rc.min_frame_bandwidth,
- FRAME_OVERHEAD_BITS);
+ rc->min_frame_bandwidth = MAX(rc->min_frame_bandwidth, FRAME_OVERHEAD_BITS);
// A maximum bitrate for a frame is defined.
// The baseline for this aligns with HW implementations that
@@ -1158,28 +1158,28 @@
// be acheived because of a user specificed max q (e.g. when the user
// specifies lossless encode.
//
- vbr_max_bits = (int)(((int64_t)cpi->rc.av_per_frame_bandwidth *
- cpi->oxcf.two_pass_vbrmax_section) / 100);
- cpi->rc.max_frame_bandwidth =
- MAX(MAX((cm->MBs * MAX_MB_RATE), MAXRATE_1080P), vbr_max_bits);
+ vbr_max_bits = (int)(((int64_t)rc->av_per_frame_bandwidth *
+ oxcf->two_pass_vbrmax_section) / 100);
+ rc->max_frame_bandwidth = MAX(MAX((cm->MBs * MAX_MB_RATE), MAXRATE_1080P),
+ vbr_max_bits);
// Set Maximum gf/arf interval
- cpi->rc.max_gf_interval = 16;
+ rc->max_gf_interval = 16;
// Extended interval for genuinely static scenes
- cpi->twopass.static_scene_max_gf_interval = cpi->key_frame_frequency >> 1;
+ rc->static_scene_max_gf_interval = cpi->key_frame_frequency >> 1;
// Special conditions when alt ref frame enabled in lagged compress mode
- if (cpi->oxcf.play_alternate && cpi->oxcf.lag_in_frames) {
- if (cpi->rc.max_gf_interval > cpi->oxcf.lag_in_frames - 1)
- cpi->rc.max_gf_interval = cpi->oxcf.lag_in_frames - 1;
+ if (oxcf->play_alternate && oxcf->lag_in_frames) {
+ if (rc->max_gf_interval > oxcf->lag_in_frames - 1)
+ rc->max_gf_interval = oxcf->lag_in_frames - 1;
- if (cpi->twopass.static_scene_max_gf_interval > cpi->oxcf.lag_in_frames - 1)
- cpi->twopass.static_scene_max_gf_interval = cpi->oxcf.lag_in_frames - 1;
+ if (rc->static_scene_max_gf_interval > oxcf->lag_in_frames - 1)
+ rc->static_scene_max_gf_interval = oxcf->lag_in_frames - 1;
}
- if (cpi->rc.max_gf_interval > cpi->twopass.static_scene_max_gf_interval)
- cpi->rc.max_gf_interval = cpi->twopass.static_scene_max_gf_interval;
+ if (rc->max_gf_interval > rc->static_scene_max_gf_interval)
+ rc->max_gf_interval = rc->static_scene_max_gf_interval;
}
static int64_t rescale(int64_t val, int64_t num, int denom) {
@@ -1410,6 +1410,10 @@
cpi->oxcf.cpu_used = clamp(cpi->oxcf.cpu_used, -5, 5);
break;
+ case MODE_BESTQUALITY:
+ cpi->pass = 0;
+ break;
+
case MODE_FIRSTPASS:
cpi->pass = 1;
break;
diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index 1540d6e..8c6b48c 100644
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -442,8 +442,6 @@
int avg_frame_size;
} LAYER_CONTEXT;
-#define MAX_SEGMENTS 8
-
typedef enum {
NORMAL = 0,
FOURFIVE = 1,
@@ -522,7 +520,7 @@
// were generated in the first encoding pass to create the compressed
// output using the highest possible quality, and taking a
// longer amount of time to encode.. ( speed setting ignored )
- int mode;
+ MODE mode;
// Key Framing Operations
int auto_key; // autodetect cut scenes and set the keyframes
@@ -533,7 +531,7 @@
// ----------------------------------------------------------------
// DATARATE CONTROL OPTIONS
- int end_usage; // vbr or cbr
+ END_USAGE end_usage; // vbr or cbr
// buffer targeting aggressiveness
int under_shoot_pct;
@@ -682,13 +680,13 @@
int rd_thresh_freq_sub8x8[BLOCK_SIZES][MAX_REFS];
int64_t rd_comp_pred_diff[REFERENCE_MODES];
- int64_t rd_prediction_type_threshes[4][REFERENCE_MODES];
+ int64_t rd_prediction_type_threshes[MAX_REF_FRAMES][REFERENCE_MODES];
int64_t rd_tx_select_diff[TX_MODES];
// FIXME(rbultje) can this overflow?
- int rd_tx_select_threshes[4][TX_MODES];
+ int rd_tx_select_threshes[MAX_REF_FRAMES][TX_MODES];
int64_t rd_filter_diff[SWITCHABLE_FILTER_CONTEXTS];
- int64_t rd_filter_threshes[4][SWITCHABLE_FILTER_CONTEXTS];
+ int64_t rd_filter_threshes[MAX_REF_FRAMES][SWITCHABLE_FILTER_CONTEXTS];
int64_t rd_filter_cache[SWITCHABLE_FILTER_CONTEXTS];
int64_t mask_filter_rd;
diff --git a/vp9/encoder/vp9_ratectrl.h b/vp9/encoder/vp9_ratectrl.h
index 5dbc7d1..a6f2c31 100644
--- a/vp9/encoder/vp9_ratectrl.h
+++ b/vp9/encoder/vp9_ratectrl.h
@@ -37,6 +37,7 @@
int frames_since_golden;
int frames_till_gf_update_due;
int max_gf_interval;
+ int static_scene_max_gf_interval;
int baseline_gf_interval;
int frames_to_key;
int frames_since_key;
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index db07927..e957b15 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -3720,16 +3720,13 @@
// combination that wins out.
if (cpi->sf.adaptive_rd_thresh) {
for (mode_index = 0; mode_index < MAX_MODES; ++mode_index) {
+ int *const fact = &cpi->rd_thresh_freq_fact[bsize][mode_index];
+
if (mode_index == best_mode_index) {
- cpi->rd_thresh_freq_fact[bsize][mode_index] -=
- (cpi->rd_thresh_freq_fact[bsize][mode_index] >> 3);
+ *fact -= (*fact >> 3);
} else {
- cpi->rd_thresh_freq_fact[bsize][mode_index] += RD_THRESH_INC;
- if (cpi->rd_thresh_freq_fact[bsize][mode_index] >
- (cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT)) {
- cpi->rd_thresh_freq_fact[bsize][mode_index] =
- cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT;
- }
+ *fact = MIN(*fact + RD_THRESH_INC,
+ cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT);
}
}
}
@@ -4445,16 +4442,13 @@
// combination that wins out.
if (cpi->sf.adaptive_rd_thresh) {
for (mode_index = 0; mode_index < MAX_REFS; ++mode_index) {
+ int *const fact = &cpi->rd_thresh_freq_sub8x8[bsize][mode_index];
+
if (mode_index == best_mode_index) {
- cpi->rd_thresh_freq_sub8x8[bsize][mode_index] -=
- (cpi->rd_thresh_freq_sub8x8[bsize][mode_index] >> 3);
+ *fact -= (*fact >> 3);
} else {
- cpi->rd_thresh_freq_sub8x8[bsize][mode_index] += RD_THRESH_INC;
- if (cpi->rd_thresh_freq_sub8x8[bsize][mode_index] >
- (cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT)) {
- cpi->rd_thresh_freq_sub8x8[bsize][mode_index] =
- cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT;
- }
+ *fact = MIN(*fact + RD_THRESH_INC,
+ cpi->sf.adaptive_rd_thresh * RD_THRESH_MAX_FACT);
}
}
}
diff --git a/vp9/vp9_cx_iface.c b/vp9/vp9_cx_iface.c
index 5ba8cdc..a20209f 100644
--- a/vp9/vp9_cx_iface.c
+++ b/vp9/vp9_cx_iface.c
@@ -577,33 +577,27 @@
}
static void pick_quickcompress_mode(vpx_codec_alg_priv_t *ctx,
- unsigned long duration,
- unsigned long deadline) {
- unsigned int new_qc;
-
- /* Use best quality mode if no deadline is given. */
- new_qc = MODE_BESTQUALITY;
+ unsigned long duration,
+ unsigned long deadline) {
+ // Use best quality mode if no deadline is given.
+ MODE new_qc = MODE_BESTQUALITY;
if (deadline) {
- uint64_t duration_us;
+ // Convert duration parameter from stream timebase to microseconds
+ const uint64_t duration_us = (uint64_t)duration * 1000000 *
+ (uint64_t)ctx->cfg.g_timebase.num /
+ (uint64_t)ctx->cfg.g_timebase.den;
- /* Convert duration parameter from stream timebase to microseconds */
- duration_us = (uint64_t)duration * 1000000
- * (uint64_t)ctx->cfg.g_timebase.num
- / (uint64_t)ctx->cfg.g_timebase.den;
-
- /* If the deadline is more that the duration this frame is to be shown,
- * use good quality mode. Otherwise use realtime mode.
- */
- new_qc = (deadline > duration_us) ? MODE_GOODQUALITY : MODE_REALTIME;
+ // If the deadline is more that the duration this frame is to be shown,
+ // use good quality mode. Otherwise use realtime mode.
+ new_qc = (deadline > duration_us) ? MODE_GOODQUALITY : MODE_REALTIME;
}
if (ctx->cfg.g_pass == VPX_RC_FIRST_PASS)
new_qc = MODE_FIRSTPASS;
else if (ctx->cfg.g_pass == VPX_RC_LAST_PASS)
- new_qc = (new_qc == MODE_BESTQUALITY)
- ? MODE_SECONDPASS_BEST
- : MODE_SECONDPASS;
+ new_qc = (new_qc == MODE_BESTQUALITY) ? MODE_SECONDPASS_BEST
+ : MODE_SECONDPASS;
if (ctx->oxcf.mode != new_qc) {
ctx->oxcf.mode = new_qc;
