From f7bd1274e38ae8ec631887841ea27a10ba366ba2 Mon Sep 17 00:00:00 2001
From: Adrian Grange <agrange@google.com>
Date: Wed, 9 Apr 2014 14:51:29 -0700
Subject: [PATCH] Enable vpxenc to specify internal coded frame size
Added command line flags "resize-width" & "resize-height"
to allow the user to specify the frame size to encode at.
These two flags are ignored if the "resize-allowed" switch
is not set to 1.
All frames in the clip are then encoded at this size, which
must be smaller than the raw frame size.
Change-Id: I3d64bd9303d5c0bd678461a866a1ea621700d744
---
vp8/vp8_cx_iface.c | 2 ++
vp9/encoder/vp9_onyx_if.c | 9 +++++++++
vp9/encoder/vp9_onyx_int.h | 5 +++++
vp9/vp9_cx_iface.c | 14 ++++++++++++++
vpx/vpx_encoder.h | 13 +++++++++++++
vpxenc.c | 18 ++++++++++++++----
6 files changed, 57 insertions(+), 4 deletions(-)
diff --git a/vp8/vp8_cx_iface.c b/vp8/vp8_cx_iface.c
index 4c896b1d14..6ca6087b27 100644
--- a/vp8/vp8_cx_iface.c
+++ b/vp8/vp8_cx_iface.c
@@ -1235,6 +1235,8 @@ static vpx_codec_enc_cfg_map_t vp8e_usage_cfg_map[] =
0, /* rc_dropframe_thresh */
0, /* rc_resize_allowed */
+ 1, /* rc_scaled_width */
+ 1, /* rc_scaled_height */
60, /* rc_resize_down_thresold */
30, /* rc_resize_up_thresold */
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index 1d70538c69..4b2aa87356 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -2997,6 +2997,15 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags,
cm->frame_flags = *frame_flags;
+ if (cpi->pass == 2 &&
+ cm->current_video_frame == 0 &&
+ cpi->oxcf.allow_spatial_resampling &&
+ cpi->oxcf.end_usage == USAGE_LOCAL_FILE_PLAYBACK) {
+ // Internal scaling is triggered on the first frame.
+ vp9_set_size_literal(cpi, cpi->oxcf.scaled_frame_width,
+ cpi->oxcf.scaled_frame_height);
+ }
+
// Reset the frame pointers to the current frame size
vp9_realloc_frame_buffer(get_frame_new_buffer(cm),
cm->width, cm->height,
diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index 9ad6ae49e4..ad56e2a853 100644
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -232,6 +232,11 @@ typedef struct VP9_CONFIG {
int lossless;
AQ_MODE aq_mode; // Adaptive Quantization mode
+ // Internal frame size scaling.
+ int allow_spatial_resampling;
+ int scaled_frame_width;
+ int scaled_frame_height;
+
// Enable feature to reduce the frame quantization every x frames.
int frame_periodic_boost;
diff --git a/vp9/vp9_cx_iface.c b/vp9/vp9_cx_iface.c
index 0623ad132b..967431c0fe 100644
--- a/vp9/vp9_cx_iface.c
+++ b/vp9/vp9_cx_iface.c
@@ -168,6 +168,11 @@ static vpx_codec_err_t validate_config(vpx_codec_alg_priv_t *ctx,
RANGE_CHECK_HI(cfg, rc_resize_down_thresh, 100);
RANGE_CHECK(cfg, g_pass, VPX_RC_ONE_PASS, VPX_RC_LAST_PASS);
+ if (cfg->rc_resize_allowed == 1) {
+ RANGE_CHECK(cfg, rc_scaled_width, 1, cfg->g_w);
+ RANGE_CHECK(cfg, rc_scaled_height, 1, cfg->g_h);
+ }
+
RANGE_CHECK(cfg, ss_number_layers, 1, VPX_SS_MAX_LAYERS);
RANGE_CHECK(cfg, ts_number_layers, 1, VPX_TS_MAX_LAYERS);
if (cfg->ts_number_layers > 1) {
@@ -332,6 +337,10 @@ static vpx_codec_err_t set_encoder_config(
oxcf->under_shoot_pct = cfg->rc_undershoot_pct;
oxcf->over_shoot_pct = cfg->rc_overshoot_pct;
+ oxcf->allow_spatial_resampling = cfg->rc_resize_allowed;
+ oxcf->scaled_frame_width = cfg->rc_scaled_width;
+ oxcf->scaled_frame_height = cfg->rc_scaled_height;
+
oxcf->maximum_buffer_size = cfg->rc_buf_sz;
oxcf->starting_buffer_level = cfg->rc_buf_initial_sz;
oxcf->optimal_buffer_level = cfg->rc_buf_optimal_sz;
@@ -410,6 +419,9 @@ static vpx_codec_err_t set_encoder_config(
printf("fixed_q: %d\n", oxcf->fixed_q);
printf("worst_allowed_q: %d\n", oxcf->worst_allowed_q);
printf("best_allowed_q: %d\n", oxcf->best_allowed_q);
+ printf("allow_spatial_resampling: %d\n", oxcf->allow_spatial_resampling);
+ printf("scaled_frame_width: %d\n", oxcf->scaled_frame_width);
+ printf("scaled_frame_height: %d\n", oxcf->scaled_frame_height);
printf("two_pass_vbrbias: %d\n", oxcf->two_pass_vbrbias);
printf("two_pass_vbrmin_section: %d\n", oxcf->two_pass_vbrmin_section);
printf("two_pass_vbrmax_section: %d\n", oxcf->two_pass_vbrmax_section);
@@ -1128,6 +1140,8 @@ static vpx_codec_enc_cfg_map_t encoder_usage_cfg_map[] = {
0, // rc_dropframe_thresh
0, // rc_resize_allowed
+ 1, // rc_scaled_width
+ 1, // rc_scaled_height
60, // rc_resize_down_thresold
30, // rc_resize_up_thresold
diff --git a/vpx/vpx_encoder.h b/vpx/vpx_encoder.h
index 2c882c19a5..571ad3fc52 100644
--- a/vpx/vpx_encoder.h
+++ b/vpx/vpx_encoder.h
@@ -396,6 +396,19 @@ extern "C" {
*/
unsigned int rc_resize_allowed;
+ /*!\brief Internal coded frame width.
+ *
+ * If spatial resampling is enabled this specifies the width of the
+ * encoded frame.
+ */
+ unsigned int rc_scaled_width;
+
+ /*!\brief Internal coded frame height.
+ *
+ * If spatial resampling is enabled this specifies the height of the
+ * encoded frame.
+ */
+ unsigned int rc_scaled_height;
/*!\brief Spatial resampling up watermark.
*
diff --git a/vpxenc.c b/vpxenc.c
index cb91d94daa..c4d642159d 100644
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -235,6 +235,10 @@ static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
"Temporal resampling threshold (buf %)");
static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
"Spatial resampling enabled (bool)");
+static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
+ "Width of encoded frame");
+static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
+ "Height of encoded frame");
static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
"Upscale threshold (buf %)");
static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
@@ -265,10 +269,10 @@ static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
"Client optimal buffer size (ms)");
static const arg_def_t *rc_args[] = {
- &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
- &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
- &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
- NULL
+ &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
+ &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
+ &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
+ &buf_initial_sz, &buf_optimal_sz, NULL
};
@@ -883,6 +887,10 @@ static int parse_stream_params(struct VpxEncoderConfig *global,
config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
} else if (arg_match(&arg, &resize_allowed, argi)) {
config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &resize_width, argi)) {
+ config->cfg.rc_scaled_width = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &resize_height, argi)) {
+ config->cfg.rc_scaled_height = arg_parse_uint(&arg);
} else if (arg_match(&arg, &resize_up_thresh, argi)) {
config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
} else if (arg_match(&arg, &resize_down_thresh, argi)) {
@@ -1067,6 +1075,8 @@ static void show_stream_config(struct stream_state *stream,
SHOW(g_lag_in_frames);
SHOW(rc_dropframe_thresh);
SHOW(rc_resize_allowed);
+ SHOW(rc_scaled_width);
+ SHOW(rc_scaled_height);
SHOW(rc_resize_up_thresh);
SHOW(rc_resize_down_thresh);
SHOW(rc_end_usage);
--
GitLab