ffmpeg/libavcodec/aacenc.c

828 lines
30 KiB
C
Raw Normal View History

/*
* AAC encoder
* Copyright (C) 2008 Konstantin Shishkov
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* AAC encoder
*/
/***********************************
* TODOs:
* add sane pulse detection
* add temporal noise shaping
***********************************/
#include "libavutil/float_dsp.h"
#include "libavutil/opt.h"
#include "avcodec.h"
#include "put_bits.h"
2012-01-28 18:28:01 +01:00
#include "internal.h"
#include "mpeg4audio.h"
#include "kbdwin.h"
#include "sinewin.h"
#include "aac.h"
#include "aactab.h"
#include "aacenc.h"
#include "aacenctab.h"
#include "aacenc_utils.h"
#include "psymodel.h"
/**
* Make AAC audio config object.
* @see 1.6.2.1 "Syntax - AudioSpecificConfig"
*/
static void put_audio_specific_config(AVCodecContext *avctx)
{
PutBitContext pb;
AACEncContext *s = avctx->priv_data;
init_put_bits(&pb, avctx->extradata, avctx->extradata_size);
put_bits(&pb, 5, 2); //object type - AAC-LC
put_bits(&pb, 4, s->samplerate_index); //sample rate index
put_bits(&pb, 4, s->channels);
//GASpecificConfig
put_bits(&pb, 1, 0); //frame length - 1024 samples
put_bits(&pb, 1, 0); //does not depend on core coder
put_bits(&pb, 1, 0); //is not extension
//Explicitly Mark SBR absent
put_bits(&pb, 11, 0x2b7); //sync extension
put_bits(&pb, 5, AOT_SBR);
put_bits(&pb, 1, 0);
flush_put_bits(&pb);
}
#define WINDOW_FUNC(type) \
static void apply_ ##type ##_window(AVFloatDSPContext *fdsp, \
SingleChannelElement *sce, \
const float *audio)
WINDOW_FUNC(only_long)
{
const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
float *out = sce->ret_buf;
fdsp->vector_fmul (out, audio, lwindow, 1024);
fdsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
}
WINDOW_FUNC(long_start)
{
const float *lwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
float *out = sce->ret_buf;
fdsp->vector_fmul(out, audio, lwindow, 1024);
memcpy(out + 1024, audio + 1024, sizeof(out[0]) * 448);
fdsp->vector_fmul_reverse(out + 1024 + 448, audio + 1024 + 448, swindow, 128);
memset(out + 1024 + 576, 0, sizeof(out[0]) * 448);
}
WINDOW_FUNC(long_stop)
{
const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
const float *swindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
float *out = sce->ret_buf;
memset(out, 0, sizeof(out[0]) * 448);
fdsp->vector_fmul(out + 448, audio + 448, swindow, 128);
memcpy(out + 576, audio + 576, sizeof(out[0]) * 448);
fdsp->vector_fmul_reverse(out + 1024, audio + 1024, lwindow, 1024);
}
WINDOW_FUNC(eight_short)
{
const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
const float *in = audio + 448;
float *out = sce->ret_buf;
int w;
for (w = 0; w < 8; w++) {
fdsp->vector_fmul (out, in, w ? pwindow : swindow, 128);
out += 128;
in += 128;
fdsp->vector_fmul_reverse(out, in, swindow, 128);
out += 128;
}
}
static void (*const apply_window[4])(AVFloatDSPContext *fdsp,
SingleChannelElement *sce,
const float *audio) = {
[ONLY_LONG_SEQUENCE] = apply_only_long_window,
[LONG_START_SEQUENCE] = apply_long_start_window,
[EIGHT_SHORT_SEQUENCE] = apply_eight_short_window,
[LONG_STOP_SEQUENCE] = apply_long_stop_window
};
static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
float *audio)
{
int i;
float *output = sce->ret_buf;
apply_window[sce->ics.window_sequence[0]](s->fdsp, sce, audio);
if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE)
s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
else
for (i = 0; i < 1024; i += 128)
s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + i, output + i*2);
memcpy(audio, audio + 1024, sizeof(audio[0]) * 1024);
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
memcpy(sce->pcoeffs, sce->coeffs, sizeof(sce->pcoeffs));
}
/**
* Encode ics_info element.
* @see Table 4.6 (syntax of ics_info)
*/
static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
{
int w;
put_bits(&s->pb, 1, 0); // ics_reserved bit
put_bits(&s->pb, 2, info->window_sequence[0]);
put_bits(&s->pb, 1, info->use_kb_window[0]);
if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
put_bits(&s->pb, 6, info->max_sfb);
put_bits(&s->pb, 1, 0); // no prediction
} else {
put_bits(&s->pb, 4, info->max_sfb);
for (w = 1; w < 8; w++)
put_bits(&s->pb, 1, !info->group_len[w]);
}
}
/**
* Encode MS data.
* @see 4.6.8.1 "Joint Coding - M/S Stereo"
*/
static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
{
int i, w;
put_bits(pb, 2, cpe->ms_mode);
if (cpe->ms_mode == 1)
for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
}
/**
* Produce integer coefficients from scalefactors provided by the model.
*/
static void adjust_frame_information(ChannelElement *cpe, int chans)
{
int i, w, w2, g, ch;
int maxsfb, cmaxsfb;
IndividualChannelStream *ics;
if (cpe->common_window) {
ics = &cpe->ch[0].ics;
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
for (w2 = 0; w2 < ics->group_len[w]; w2++) {
int start = (w+w2) * 128;
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
for (g = 0; g < ics->num_swb; g++) {
//apply Intensity stereo coeffs transformation
if (cpe->is_mask[w*16 + g]) {
int p = -1 + 2 * (cpe->ch[1].band_type[w*16+g] - 14);
float scale = cpe->ch[0].is_ener[w*16+g];
for (i = 0; i < ics->swb_sizes[g]; i++) {
cpe->ch[0].coeffs[start+i] = (cpe->ch[0].pcoeffs[start+i] + p*cpe->ch[1].pcoeffs[start+i]) * scale;
cpe->ch[1].coeffs[start+i] = 0.0f;
}
} else if (cpe->ms_mask[w*16 + g] &&
cpe->ch[0].band_type[w*16 + g] < NOISE_BT &&
cpe->ch[1].band_type[w*16 + g] < NOISE_BT) {
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
for (i = 0; i < ics->swb_sizes[g]; i++) {
cpe->ch[0].coeffs[start+i] = (cpe->ch[0].pcoeffs[start+i] + cpe->ch[1].pcoeffs[start+i]) * 0.5f;
cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].pcoeffs[start+i];
}
}
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
start += ics->swb_sizes[g];
}
}
}
}
for (ch = 0; ch < chans; ch++) {
IndividualChannelStream *ics = &cpe->ch[ch].ics;
maxsfb = 0;
cpe->ch[ch].pulse.num_pulse = 0;
for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
for (w2 = 0; w2 < ics->group_len[w]; w2++) {
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w*16+cmaxsfb-1]; cmaxsfb--)
;
maxsfb = FFMAX(maxsfb, cmaxsfb);
}
}
ics->max_sfb = maxsfb;
//adjust zero bands for window groups
for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
for (g = 0; g < ics->max_sfb; g++) {
i = 1;
for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
if (!cpe->ch[ch].zeroes[w2*16 + g]) {
i = 0;
break;
}
}
cpe->ch[ch].zeroes[w*16 + g] = i;
}
}
}
if (chans > 1 && cpe->common_window) {
IndividualChannelStream *ics0 = &cpe->ch[0].ics;
IndividualChannelStream *ics1 = &cpe->ch[1].ics;
int msc = 0;
ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
ics1->max_sfb = ics0->max_sfb;
for (w = 0; w < ics0->num_windows*16; w += 16)
for (i = 0; i < ics0->max_sfb; i++)
if (cpe->ms_mask[w+i])
msc++;
if (msc == 0 || ics0->max_sfb == 0)
cpe->ms_mode = 0;
else
cpe->ms_mode = msc < ics0->max_sfb * ics0->num_windows ? 1 : 2;
}
}
/**
* Encode scalefactor band coding type.
*/
static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
{
int w;
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
}
/**
* Encode scalefactors.
*/
static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
SingleChannelElement *sce)
{
int diff, off_sf = sce->sf_idx[0], off_pns = sce->sf_idx[0] - NOISE_OFFSET;
int off_is = 0, noise_flag = 1;
int i, w;
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
for (i = 0; i < sce->ics.max_sfb; i++) {
if (!sce->zeroes[w*16 + i]) {
if (sce->band_type[w*16 + i] == NOISE_BT) {
diff = sce->sf_idx[w*16 + i] - off_pns;
off_pns = sce->sf_idx[w*16 + i];
if (noise_flag-- > 0) {
put_bits(&s->pb, NOISE_PRE_BITS, diff + NOISE_PRE);
continue;
}
} else if (sce->band_type[w*16 + i] == INTENSITY_BT ||
sce->band_type[w*16 + i] == INTENSITY_BT2) {
diff = sce->sf_idx[w*16 + i] - off_is;
off_is = sce->sf_idx[w*16 + i];
} else {
diff = sce->sf_idx[w*16 + i] - off_sf;
off_sf = sce->sf_idx[w*16 + i];
}
diff += SCALE_DIFF_ZERO;
av_assert0(diff >= 0 && diff <= 120);
put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
}
}
}
}
/**
* Encode pulse data.
*/
static void encode_pulses(AACEncContext *s, Pulse *pulse)
{
int i;
put_bits(&s->pb, 1, !!pulse->num_pulse);
if (!pulse->num_pulse)
return;
put_bits(&s->pb, 2, pulse->num_pulse - 1);
put_bits(&s->pb, 6, pulse->start);
for (i = 0; i < pulse->num_pulse; i++) {
put_bits(&s->pb, 5, pulse->pos[i]);
put_bits(&s->pb, 4, pulse->amp[i]);
}
}
/**
* Encode spectral coefficients processed by psychoacoustic model.
*/
static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
{
int start, i, w, w2;
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
start = 0;
for (i = 0; i < sce->ics.max_sfb; i++) {
if (sce->zeroes[w*16 + i]) {
start += sce->ics.swb_sizes[i];
continue;
}
for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
sce->ics.swb_sizes[i],
sce->sf_idx[w*16 + i],
sce->band_type[w*16 + i],
s->lambda, sce->ics.window_clipping[w]);
start += sce->ics.swb_sizes[i];
}
}
}
/**
* Downscale spectral coefficients for near-clipping windows to avoid artifacts
*/
static void avoid_clipping(AACEncContext *s, SingleChannelElement *sce)
{
int start, i, j, w;
if (sce->ics.clip_avoidance_factor < 1.0f) {
for (w = 0; w < sce->ics.num_windows; w++) {
start = 0;
for (i = 0; i < sce->ics.max_sfb; i++) {
float *swb_coeffs = sce->coeffs + start + w*128;
for (j = 0; j < sce->ics.swb_sizes[i]; j++)
swb_coeffs[j] *= sce->ics.clip_avoidance_factor;
start += sce->ics.swb_sizes[i];
}
}
}
}
/**
* Encode one channel of audio data.
*/
static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
SingleChannelElement *sce,
int common_window)
{
put_bits(&s->pb, 8, sce->sf_idx[0]);
if (!common_window)
put_ics_info(s, &sce->ics);
encode_band_info(s, sce);
encode_scale_factors(avctx, s, sce);
encode_pulses(s, &sce->pulse);
put_bits(&s->pb, 1, 0); //tns
put_bits(&s->pb, 1, 0); //ssr
encode_spectral_coeffs(s, sce);
return 0;
}
/**
* Write some auxiliary information about the created AAC file.
*/
static void put_bitstream_info(AACEncContext *s, const char *name)
{
int i, namelen, padbits;
namelen = strlen(name) + 2;
put_bits(&s->pb, 3, TYPE_FIL);
put_bits(&s->pb, 4, FFMIN(namelen, 15));
if (namelen >= 15)
put_bits(&s->pb, 8, namelen - 14);
put_bits(&s->pb, 4, 0); //extension type - filler
padbits = -put_bits_count(&s->pb) & 7;
avpriv_align_put_bits(&s->pb);
for (i = 0; i < namelen - 2; i++)
put_bits(&s->pb, 8, name[i]);
put_bits(&s->pb, 12 - padbits, 0);
}
/*
2012-08-25 19:04:33 +02:00
* Copy input samples.
* Channels are reordered from libavcodec's default order to AAC order.
*/
2012-08-25 19:04:33 +02:00
static void copy_input_samples(AACEncContext *s, const AVFrame *frame)
{
2012-08-25 19:04:33 +02:00
int ch;
int end = 2048 + (frame ? frame->nb_samples : 0);
const uint8_t *channel_map = aac_chan_maps[s->channels - 1];
2012-08-25 19:04:33 +02:00
/* copy and remap input samples */
for (ch = 0; ch < s->channels; ch++) {
/* copy last 1024 samples of previous frame to the start of the current frame */
memcpy(&s->planar_samples[ch][1024], &s->planar_samples[ch][2048], 1024 * sizeof(s->planar_samples[0][0]));
2012-08-25 19:04:33 +02:00
/* copy new samples and zero any remaining samples */
2012-01-28 18:28:01 +01:00
if (frame) {
2012-08-25 19:04:33 +02:00
memcpy(&s->planar_samples[ch][2048],
frame->extended_data[channel_map[ch]],
frame->nb_samples * sizeof(s->planar_samples[0][0]));
}
2012-08-25 19:04:33 +02:00
memset(&s->planar_samples[ch][end], 0,
(3072 - end) * sizeof(s->planar_samples[0][0]));
}
}
2012-01-28 18:28:01 +01:00
static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
const AVFrame *frame, int *got_packet_ptr)
{
AACEncContext *s = avctx->priv_data;
float **samples = s->planar_samples, *samples2, *la, *overlap;
ChannelElement *cpe;
aacenc: implement Intensity Stereo encoding support This commit implements intensity stereo coding support to the native aac encoder. This is a way to increase the efficiency of the encoder by zeroing the right channel's spectral coefficients (in a channel pair) and rederiving them in the decoder using information from the scalefactor indices of special band types. This commit confomrs to the official ISO 13818-7 specifications, although due to their ambiguity certain deviations have been taken to ensure maximum sound quality. This commit has been extensively tested and has shown to not result in audiable audio artifacts unless in extreme cases. This commit also adds an option, aac_is, which has the value of 0 by default. Intensity Stereo is part of the scalable aac profile and is thus non-default. The way IS coding works is that it rederives the right channel's spectral coefficients from the left channel via the scalefactor index values left in the right channel. Since an entire band's spectral coefficients do not need to be coded, the encoder's efficiency jumps up and it unzeroes some high frequency values which it previously did not have enough bits to encode. That way less information is lost than the information lost by rederiving the spectral coefficients with some error. This is why the filesize of files encoded with IS do not decrease significantly. Users wishing that IS coding should reduce filesize are expected to reduce their encoding bitrates appropriately. This is V2 of the commit. The old version did not mark ms_mask as 0 since M/S and IS coding are incompactible, which resulted in distortions with M/S coding enabled. This version also improves phase detection by measuring it for every spectral coefficient in the band and using a simple majority rule to determine whether the coefficients are in or out of phase. Also, the energy values per spectral coefficient were changed as to reflect the official specifications. Reviewed-by: Claudio Freire <klaussfreire@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-07-02 20:13:07 +02:00
int i, ch, w, g, chans, tag, start_ch, ret, ms_mode = 0, is_mode = 0;
int chan_el_counter[4];
FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
if (s->last_frame == 2)
return 0;
2012-01-28 18:28:01 +01:00
/* add current frame to queue */
if (frame) {
if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
2012-01-28 18:28:01 +01:00
return ret;
}
2012-08-25 19:04:33 +02:00
copy_input_samples(s, frame);
if (s->psypp)
ff_psy_preprocess(s->psypp, s->planar_samples, s->channels);
if (!avctx->frame_number)
return 0;
start_ch = 0;
for (i = 0; i < s->chan_map[0]; i++) {
FFPsyWindowInfo* wi = windows + start_ch;
tag = s->chan_map[i+1];
chans = tag == TYPE_CPE ? 2 : 1;
cpe = &s->cpe[i];
for (ch = 0; ch < chans; ch++) {
IndividualChannelStream *ics = &cpe->ch[ch].ics;
int cur_channel = start_ch + ch;
float clip_avoidance_factor;
overlap = &samples[cur_channel][0];
samples2 = overlap + 1024;
la = samples2 + (448+64);
2012-01-28 18:28:01 +01:00
if (!frame)
la = NULL;
if (tag == TYPE_LFE) {
wi[ch].window_type[0] = ONLY_LONG_SEQUENCE;
wi[ch].window_shape = 0;
wi[ch].num_windows = 1;
wi[ch].grouping[0] = 1;
/* Only the lowest 12 coefficients are used in a LFE channel.
* The expression below results in only the bottom 8 coefficients
* being used for 11.025kHz to 16kHz sample rates.
*/
ics->num_swb = s->samplerate_index >= 8 ? 1 : 3;
} else {
wi[ch] = s->psy.model->window(&s->psy, samples2, la, cur_channel,
ics->window_sequence[0]);
}
ics->window_sequence[1] = ics->window_sequence[0];
ics->window_sequence[0] = wi[ch].window_type[0];
ics->use_kb_window[1] = ics->use_kb_window[0];
ics->use_kb_window[0] = wi[ch].window_shape;
ics->num_windows = wi[ch].num_windows;
ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
ics->num_swb = tag == TYPE_LFE ? ics->num_swb : s->psy.num_bands[ics->num_windows == 8];
clip_avoidance_factor = 0.0f;
for (w = 0; w < ics->num_windows; w++)
ics->group_len[w] = wi[ch].grouping[w];
for (w = 0; w < ics->num_windows; w++) {
if (wi[ch].clipping[w] > CLIP_AVOIDANCE_FACTOR) {
ics->window_clipping[w] = 1;
clip_avoidance_factor = FFMAX(clip_avoidance_factor, wi[ch].clipping[w]);
} else {
ics->window_clipping[w] = 0;
}
}
if (clip_avoidance_factor > CLIP_AVOIDANCE_FACTOR) {
ics->clip_avoidance_factor = CLIP_AVOIDANCE_FACTOR / clip_avoidance_factor;
} else {
ics->clip_avoidance_factor = 1.0f;
}
apply_window_and_mdct(s, &cpe->ch[ch], overlap);
if (isnan(cpe->ch->coeffs[0])) {
av_log(avctx, AV_LOG_ERROR, "Input contains NaN\n");
return AVERROR(EINVAL);
}
avoid_clipping(s, &cpe->ch[ch]);
}
start_ch += chans;
}
if ((ret = ff_alloc_packet2(avctx, avpkt, 8192 * s->channels, 0)) < 0)
return ret;
do {
int frame_bits;
2012-01-28 18:28:01 +01:00
init_put_bits(&s->pb, avpkt->data, avpkt->size);
if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & AV_CODEC_FLAG_BITEXACT))
put_bitstream_info(s, LIBAVCODEC_IDENT);
start_ch = 0;
memset(chan_el_counter, 0, sizeof(chan_el_counter));
for (i = 0; i < s->chan_map[0]; i++) {
FFPsyWindowInfo* wi = windows + start_ch;
const float *coeffs[2];
tag = s->chan_map[i+1];
chans = tag == TYPE_CPE ? 2 : 1;
cpe = &s->cpe[i];
memset(cpe->is_mask, 0, sizeof(cpe->is_mask));
memset(cpe->ms_mask, 0, sizeof(cpe->ms_mask));
put_bits(&s->pb, 3, tag);
put_bits(&s->pb, 4, chan_el_counter[tag]++);
for (ch = 0; ch < chans; ch++)
coeffs[ch] = cpe->ch[ch].coeffs;
s->psy.model->analyze(&s->psy, start_ch, coeffs, wi);
for (ch = 0; ch < chans; ch++) {
s->cur_channel = start_ch + ch;
s->coder->search_for_quantizers(avctx, s, &cpe->ch[ch], s->lambda);
}
cpe->common_window = 0;
if (chans > 1
&& wi[0].window_type[0] == wi[1].window_type[0]
&& wi[0].window_shape == wi[1].window_shape) {
cpe->common_window = 1;
for (w = 0; w < wi[0].num_windows; w++) {
if (wi[0].grouping[w] != wi[1].grouping[w]) {
cpe->common_window = 0;
break;
}
}
}
if (s->options.pns && s->coder->search_for_pns) {
for (ch = 0; ch < chans; ch++) {
s->cur_channel = start_ch + ch;
s->coder->search_for_pns(s, avctx, &cpe->ch[ch]);
}
}
s->cur_channel = start_ch;
if (s->options.stereo_mode && cpe->common_window) {
if (s->options.stereo_mode > 0) {
IndividualChannelStream *ics = &cpe->ch[0].ics;
for (w = 0; w < ics->num_windows; w += ics->group_len[w])
for (g = 0; g < ics->num_swb; g++)
cpe->ms_mask[w*16+g] = 1;
} else if (s->coder->search_for_ms) {
s->coder->search_for_ms(s, cpe);
}
}
aacenc: implement Intensity Stereo encoding support This commit implements intensity stereo coding support to the native aac encoder. This is a way to increase the efficiency of the encoder by zeroing the right channel's spectral coefficients (in a channel pair) and rederiving them in the decoder using information from the scalefactor indices of special band types. This commit confomrs to the official ISO 13818-7 specifications, although due to their ambiguity certain deviations have been taken to ensure maximum sound quality. This commit has been extensively tested and has shown to not result in audiable audio artifacts unless in extreme cases. This commit also adds an option, aac_is, which has the value of 0 by default. Intensity Stereo is part of the scalable aac profile and is thus non-default. The way IS coding works is that it rederives the right channel's spectral coefficients from the left channel via the scalefactor index values left in the right channel. Since an entire band's spectral coefficients do not need to be coded, the encoder's efficiency jumps up and it unzeroes some high frequency values which it previously did not have enough bits to encode. That way less information is lost than the information lost by rederiving the spectral coefficients with some error. This is why the filesize of files encoded with IS do not decrease significantly. Users wishing that IS coding should reduce filesize are expected to reduce their encoding bitrates appropriately. This is V2 of the commit. The old version did not mark ms_mask as 0 since M/S and IS coding are incompactible, which resulted in distortions with M/S coding enabled. This version also improves phase detection by measuring it for every spectral coefficient in the band and using a simple majority rule to determine whether the coefficients are in or out of phase. Also, the energy values per spectral coefficient were changed as to reflect the official specifications. Reviewed-by: Claudio Freire <klaussfreire@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-07-02 20:13:07 +02:00
if (chans > 1 && s->options.intensity_stereo && s->coder->search_for_is) {
s->coder->search_for_is(s, avctx, cpe);
aacenc: implement Intensity Stereo encoding support This commit implements intensity stereo coding support to the native aac encoder. This is a way to increase the efficiency of the encoder by zeroing the right channel's spectral coefficients (in a channel pair) and rederiving them in the decoder using information from the scalefactor indices of special band types. This commit confomrs to the official ISO 13818-7 specifications, although due to their ambiguity certain deviations have been taken to ensure maximum sound quality. This commit has been extensively tested and has shown to not result in audiable audio artifacts unless in extreme cases. This commit also adds an option, aac_is, which has the value of 0 by default. Intensity Stereo is part of the scalable aac profile and is thus non-default. The way IS coding works is that it rederives the right channel's spectral coefficients from the left channel via the scalefactor index values left in the right channel. Since an entire band's spectral coefficients do not need to be coded, the encoder's efficiency jumps up and it unzeroes some high frequency values which it previously did not have enough bits to encode. That way less information is lost than the information lost by rederiving the spectral coefficients with some error. This is why the filesize of files encoded with IS do not decrease significantly. Users wishing that IS coding should reduce filesize are expected to reduce their encoding bitrates appropriately. This is V2 of the commit. The old version did not mark ms_mask as 0 since M/S and IS coding are incompactible, which resulted in distortions with M/S coding enabled. This version also improves phase detection by measuring it for every spectral coefficient in the band and using a simple majority rule to determine whether the coefficients are in or out of phase. Also, the energy values per spectral coefficient were changed as to reflect the official specifications. Reviewed-by: Claudio Freire <klaussfreire@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-07-02 20:13:07 +02:00
if (cpe->is_mode) is_mode = 1;
}
if (s->coder->set_special_band_scalefactors)
for (ch = 0; ch < chans; ch++)
s->coder->set_special_band_scalefactors(s, &cpe->ch[ch]);
adjust_frame_information(cpe, chans);
if (chans == 2) {
put_bits(&s->pb, 1, cpe->common_window);
if (cpe->common_window) {
put_ics_info(s, &cpe->ch[0].ics);
encode_ms_info(&s->pb, cpe);
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
if (cpe->ms_mode) ms_mode = 1;
}
}
for (ch = 0; ch < chans; ch++) {
s->cur_channel = start_ch + ch;
encode_individual_channel(avctx, s, &cpe->ch[ch], cpe->common_window);
}
start_ch += chans;
}
frame_bits = put_bits_count(&s->pb);
if (frame_bits <= 6144 * s->channels - 3) {
s->psy.bitres.bits = frame_bits / s->channels;
break;
}
aacenc: implement Intensity Stereo encoding support This commit implements intensity stereo coding support to the native aac encoder. This is a way to increase the efficiency of the encoder by zeroing the right channel's spectral coefficients (in a channel pair) and rederiving them in the decoder using information from the scalefactor indices of special band types. This commit confomrs to the official ISO 13818-7 specifications, although due to their ambiguity certain deviations have been taken to ensure maximum sound quality. This commit has been extensively tested and has shown to not result in audiable audio artifacts unless in extreme cases. This commit also adds an option, aac_is, which has the value of 0 by default. Intensity Stereo is part of the scalable aac profile and is thus non-default. The way IS coding works is that it rederives the right channel's spectral coefficients from the left channel via the scalefactor index values left in the right channel. Since an entire band's spectral coefficients do not need to be coded, the encoder's efficiency jumps up and it unzeroes some high frequency values which it previously did not have enough bits to encode. That way less information is lost than the information lost by rederiving the spectral coefficients with some error. This is why the filesize of files encoded with IS do not decrease significantly. Users wishing that IS coding should reduce filesize are expected to reduce their encoding bitrates appropriately. This is V2 of the commit. The old version did not mark ms_mask as 0 since M/S and IS coding are incompactible, which resulted in distortions with M/S coding enabled. This version also improves phase detection by measuring it for every spectral coefficient in the band and using a simple majority rule to determine whether the coefficients are in or out of phase. Also, the energy values per spectral coefficient were changed as to reflect the official specifications. Reviewed-by: Claudio Freire <klaussfreire@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-07-02 20:13:07 +02:00
if (is_mode || ms_mode) {
AAC: Fix M/S stereo encoding This patch fixes a pointer arithmetic bug in adjust_frame_information that resulted in heavily corrupted audio when using M/S encoding. Also, a backup copy of untransformed coefficients has to be kept around or attempts at re-processing the frame (which happens when hevavily overspending bits during transients) will result in re-encoding of the coefficients and subsequent corruption of the resulting stream. A/B testing shows the bug as corrected, but still cannot prove that M/S coding is a win at least in numbers. Limited listening tests do show improvement on M/S encoded samples in lower bitrates, but they're hidden among the other artifacts that remain to be corrected in the encoder. Some of the regressions flagged in the report do show poor stereo image (but not buggy), so M/S encoding is clearly not good enough yet to be defaulted to auto. In numbers, Patched against Unpatched, stereo_mode auto: Files: 114 Bitrates: 6 Tests: 683 Serious Regressions: 0 (0%) Regressions: 0 (0%) Improvements: 227 (33%) Big improvements: 92 (13%) Worst regression - mybloodrusts.wv - 256k - StdDev: 28.61 pSNR: -0.43 maxdiff: 1372.00 Best improvement - 60.wv - 384k - StdDev: -369.57 pSNR: 45.02 maxdiff: -13322.00 Average - StdDev: -80.56 pSNR: 2.49 maxdiff: -8858.00 Patched against Unpatched stereo_mode ms_off shows no difference. Patched stereo_mode auto vs Unpatched stereo_mode ms_off shows a small average improvement, just not too significant: Serious Regressions: 0 (0%) Regressions: 10 (1%) Improvements: 45 (6%) Big improvements: 2 (0%) Worst regression - Illinois.wv - 256k - StdDev: 33.20 pSNR: -2.03 maxdiff: 477.00 Best improvement - song_of_circomstances.flac - 384k - StdDev: -3.97 pSNR: 7.61 maxdiff: -826.00 Average - StdDev: -10.25 pSNR: 0.20 maxdiff: -281.00 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-03-03 07:43:06 +01:00
for (i = 0; i < s->chan_map[0]; i++) {
// Must restore coeffs
chans = tag == TYPE_CPE ? 2 : 1;
cpe = &s->cpe[i];
for (ch = 0; ch < chans; ch++)
memcpy(cpe->ch[ch].coeffs, cpe->ch[ch].pcoeffs, sizeof(cpe->ch[ch].coeffs));
}
}
s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
} while (1);
put_bits(&s->pb, 3, TYPE_END);
flush_put_bits(&s->pb);
avctx->frame_bits = put_bits_count(&s->pb);
// rate control stuff
if (!(avctx->flags & AV_CODEC_FLAG_QSCALE)) {
float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
s->lambda *= ratio;
s->lambda = FFMIN(s->lambda, 65536.f);
}
2012-01-28 18:28:01 +01:00
if (!frame)
s->last_frame++;
2012-01-28 18:28:01 +01:00
ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
&avpkt->duration);
avpkt->size = put_bits_count(&s->pb) >> 3;
*got_packet_ptr = 1;
return 0;
}
static av_cold int aac_encode_end(AVCodecContext *avctx)
{
AACEncContext *s = avctx->priv_data;
ff_mdct_end(&s->mdct1024);
ff_mdct_end(&s->mdct128);
ff_psy_end(&s->psy);
if (s->psypp)
ff_psy_preprocess_end(s->psypp);
av_freep(&s->buffer.samples);
av_freep(&s->cpe);
av_freep(&s->fdsp);
2012-01-28 18:28:01 +01:00
ff_af_queue_close(&s->afq);
return 0;
}
static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
{
int ret = 0;
s->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
if (!s->fdsp)
return AVERROR(ENOMEM);
// window init
ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
ff_init_ff_sine_windows(10);
ff_init_ff_sine_windows(7);
if ((ret = ff_mdct_init(&s->mdct1024, 11, 0, 32768.0)) < 0)
return ret;
if ((ret = ff_mdct_init(&s->mdct128, 8, 0, 32768.0)) < 0)
return ret;
return 0;
}
static av_cold int alloc_buffers(AVCodecContext *avctx, AACEncContext *s)
{
int ch;
FF_ALLOCZ_ARRAY_OR_GOTO(avctx, s->buffer.samples, s->channels, 3 * 1024 * sizeof(s->buffer.samples[0]), alloc_fail);
FF_ALLOCZ_ARRAY_OR_GOTO(avctx, s->cpe, s->chan_map[0], sizeof(ChannelElement), alloc_fail);
FF_ALLOCZ_OR_GOTO(avctx, avctx->extradata, 5 + AV_INPUT_BUFFER_PADDING_SIZE, alloc_fail);
for(ch = 0; ch < s->channels; ch++)
s->planar_samples[ch] = s->buffer.samples + 3 * 1024 * ch;
return 0;
alloc_fail:
return AVERROR(ENOMEM);
}
static av_cold int aac_encode_init(AVCodecContext *avctx)
{
AACEncContext *s = avctx->priv_data;
int i, ret = 0;
const uint8_t *sizes[2];
uint8_t grouping[AAC_MAX_CHANNELS];
int lengths[2];
avctx->frame_size = 1024;
for (i = 0; i < 16; i++)
if (avctx->sample_rate == avpriv_mpeg4audio_sample_rates[i])
break;
s->channels = avctx->channels;
ERROR_IF(i == 16 || i >= swb_size_1024_len || i >= swb_size_128_len,
"Unsupported sample rate %d\n", avctx->sample_rate);
ERROR_IF(s->channels > AAC_MAX_CHANNELS,
"Unsupported number of channels: %d\n", s->channels);
ERROR_IF(avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW,
"Unsupported profile %d\n", avctx->profile);
WARN_IF(1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * s->channels,
"Too many bits per frame requested, clamping to max\n");
avctx->bit_rate = (int)FFMIN(
6144 * s->channels / 1024.0 * avctx->sample_rate,
avctx->bit_rate);
s->samplerate_index = i;
s->chan_map = aac_chan_configs[s->channels-1];
if ((ret = dsp_init(avctx, s)) < 0)
goto fail;
if ((ret = alloc_buffers(avctx, s)) < 0)
goto fail;
avctx->extradata_size = 5;
put_audio_specific_config(avctx);
sizes[0] = swb_size_1024[i];
sizes[1] = swb_size_128[i];
lengths[0] = ff_aac_num_swb_1024[i];
lengths[1] = ff_aac_num_swb_128[i];
for (i = 0; i < s->chan_map[0]; i++)
grouping[i] = s->chan_map[i + 1] == TYPE_CPE;
if ((ret = ff_psy_init(&s->psy, avctx, 2, sizes, lengths,
s->chan_map[0], grouping)) < 0)
goto fail;
s->psypp = ff_psy_preprocess_init(avctx);
s->coder = &ff_aac_coders[s->options.aac_coder];
if (HAVE_MIPSDSPR1)
ff_aac_coder_init_mips(s);
s->lambda = avctx->global_quality > 0 ? avctx->global_quality : 120;
ff_aac_tableinit();
avctx->initial_padding = 1024;
2012-01-28 18:28:01 +01:00
ff_af_queue_init(avctx, &s->afq);
return 0;
fail:
aac_encode_end(avctx);
return ret;
}
#define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
static const AVOption aacenc_options[] = {
{"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), AV_OPT_TYPE_INT, {.i64 = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
{"auto", "Selected by the Encoder", 0, AV_OPT_TYPE_CONST, {.i64 = -1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
{"ms_off", "Disable Mid/Side coding", 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
{"ms_force", "Force Mid/Side for the whole frame if possible", 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
{"aac_coder", "Coding algorithm", offsetof(AACEncContext, options.aac_coder), AV_OPT_TYPE_INT, {.i64 = AAC_CODER_TWOLOOP}, 0, AAC_CODER_NB-1, AACENC_FLAGS, "aac_coder"},
{"faac", "FAAC-inspired method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAAC}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
{"anmr", "ANMR method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_ANMR}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
{"twoloop", "Two loop searching method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_TWOLOOP}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
{"fast", "Constant quantizer", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAST}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
aaccoder: Implement Perceptual Noise Substitution for AAC This commit implements the perceptual noise substitution AAC extension. This is a proof of concept implementation, and as such, is not enabled by default. This is the fourth revision of this patch, made after some problems were noted out. Any changes made since the previous revisions have been indicated. In order to extend the encoder to use an additional codebook, the array holding each codebook has been modified with two additional entries - 13 for the NOISE_BT codebook and 12 which has a placeholder function. The cost system was modified to skip the 12th entry using an array to map the input and outputs it has. It also does not accept using the 13th codebook for any band which is not marked as containing noise, thereby restricting its ability to arbitrarily choose it for bands. The use of arrays allows the system to be easily extended to allow for intensity stereo encoding, which uses additional codebooks. The 12th entry in the codebook function array points to a function which stops the execution of the program by calling an assert with an always 'false' argument. It was pointed out in an email discussion with Claudio Freire that having a 'NULL' entry can result in unexpected behaviour and could be used as a security hole. There is no danger of this function being called during encoding due to the codebook maps introduced. Another change from version 1 of the patch is the addition of an argument to the encoder, '-aac_pns' to enable and disable the PNS. This currently defaults to disable the PNS, as it is experimental. The switch will be removed in the future, when the algorithm to select noise bands has been improved. The current algorithm simply compares the energy to the threshold (multiplied by a constant) to determine noise, however the FFPsyBand structure contains other useful figures to determine which bands carry noise more accurately. Some of the sample files provided triggered an assertion when the parameter to tune the threshold was set to a value of '2.2'. Claudio Freire reported the problem's source could be in the range of the scalefactor indices for noise and advised to measure the minimal index and clip anything above the maximum allowed value. This has been implemented and all the files which used to trigger the asserion now encode without error. The third revision of the problem also removes unneded variabes and comparisons. All of them were redundant and were of little use for when the PNS implementation would be extended. The fourth revision moved the clipping of the noise scalefactors outside the second loop of the two-loop algorithm in order to prevent their redundant calculations. Also, freq_mult has been changed to a float variable due to the fact that rounding errors can prove to be a problem at low frequencies. Considerations were taken whether the entire expression could be evaluated inside the expression , but in the end it was decided that it would be for the best if just the type of the variable were to change. Claudio Freire reported the two problems. There is no change of functionality (except for low sampling frequencies) so the spectral demonstrations at the end of this commit's message were not updated. Finally, the way energy values are converted to scalefactor indices has changed since the first commit, as per the suggestion of Claudio Freire. This may still have some drawbacks, but unlike the first commit it works without having redundant offsets and outputs what the decoder expects to have, in terms of the ranges of the scalefactor indices. Some spectral comparisons: https://trac.ffmpeg.org/attachment/wiki/Encode/AAC/Original.png (original), https://trac.ffmpeg.org/attachment/wiki/Encode/AAC/PNS_NO.png (encoded without PNS), https://trac.ffmpeg.org/attachment/wiki/Encode/AAC/PNS1.2.png (encoded with PNS, const = 1.2), https://trac.ffmpeg.org/attachment/wiki/Encode/AAC/Difference1.png (spectral difference). The constant is the value which multiplies the threshold when it gets compared to the energy, larger values means more noise will be substituded by PNS values. Example when const = 2.2: https://trac.ffmpeg.org/attachment/wiki/Encode/AAC/PNS_2.2.png Reviewed-by: Claudio Freire <klaussfreire@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-04-15 13:18:42 +02:00
{"aac_pns", "Perceptual Noise Substitution", offsetof(AACEncContext, options.pns), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AACENC_FLAGS, "aac_pns"},
{"disable", "Disable perceptual noise substitution", 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_pns"},
{"enable", "Enable perceptual noise substitution", 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_pns"},
aacenc: implement Intensity Stereo encoding support This commit implements intensity stereo coding support to the native aac encoder. This is a way to increase the efficiency of the encoder by zeroing the right channel's spectral coefficients (in a channel pair) and rederiving them in the decoder using information from the scalefactor indices of special band types. This commit confomrs to the official ISO 13818-7 specifications, although due to their ambiguity certain deviations have been taken to ensure maximum sound quality. This commit has been extensively tested and has shown to not result in audiable audio artifacts unless in extreme cases. This commit also adds an option, aac_is, which has the value of 0 by default. Intensity Stereo is part of the scalable aac profile and is thus non-default. The way IS coding works is that it rederives the right channel's spectral coefficients from the left channel via the scalefactor index values left in the right channel. Since an entire band's spectral coefficients do not need to be coded, the encoder's efficiency jumps up and it unzeroes some high frequency values which it previously did not have enough bits to encode. That way less information is lost than the information lost by rederiving the spectral coefficients with some error. This is why the filesize of files encoded with IS do not decrease significantly. Users wishing that IS coding should reduce filesize are expected to reduce their encoding bitrates appropriately. This is V2 of the commit. The old version did not mark ms_mask as 0 since M/S and IS coding are incompactible, which resulted in distortions with M/S coding enabled. This version also improves phase detection by measuring it for every spectral coefficient in the band and using a simple majority rule to determine whether the coefficients are in or out of phase. Also, the energy values per spectral coefficient were changed as to reflect the official specifications. Reviewed-by: Claudio Freire <klaussfreire@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-07-02 20:13:07 +02:00
{"aac_is", "Intensity stereo coding", offsetof(AACEncContext, options.intensity_stereo), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AACENC_FLAGS, "intensity_stereo"},
{"disable", "Disable intensity stereo coding", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, INT_MIN, INT_MAX, AACENC_FLAGS, "intensity_stereo"},
aacenc: implement Intensity Stereo encoding support This commit implements intensity stereo coding support to the native aac encoder. This is a way to increase the efficiency of the encoder by zeroing the right channel's spectral coefficients (in a channel pair) and rederiving them in the decoder using information from the scalefactor indices of special band types. This commit confomrs to the official ISO 13818-7 specifications, although due to their ambiguity certain deviations have been taken to ensure maximum sound quality. This commit has been extensively tested and has shown to not result in audiable audio artifacts unless in extreme cases. This commit also adds an option, aac_is, which has the value of 0 by default. Intensity Stereo is part of the scalable aac profile and is thus non-default. The way IS coding works is that it rederives the right channel's spectral coefficients from the left channel via the scalefactor index values left in the right channel. Since an entire band's spectral coefficients do not need to be coded, the encoder's efficiency jumps up and it unzeroes some high frequency values which it previously did not have enough bits to encode. That way less information is lost than the information lost by rederiving the spectral coefficients with some error. This is why the filesize of files encoded with IS do not decrease significantly. Users wishing that IS coding should reduce filesize are expected to reduce their encoding bitrates appropriately. This is V2 of the commit. The old version did not mark ms_mask as 0 since M/S and IS coding are incompactible, which resulted in distortions with M/S coding enabled. This version also improves phase detection by measuring it for every spectral coefficient in the band and using a simple majority rule to determine whether the coefficients are in or out of phase. Also, the energy values per spectral coefficient were changed as to reflect the official specifications. Reviewed-by: Claudio Freire <klaussfreire@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2015-07-02 20:13:07 +02:00
{"enable", "Enable intensity stereo coding", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, INT_MIN, INT_MAX, AACENC_FLAGS, "intensity_stereo"},
{NULL}
};
static const AVClass aacenc_class = {
"AAC encoder",
av_default_item_name,
aacenc_options,
LIBAVUTIL_VERSION_INT,
};
AVCodec ff_aac_encoder = {
.name = "aac",
.long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
.type = AVMEDIA_TYPE_AUDIO,
.id = AV_CODEC_ID_AAC,
.priv_data_size = sizeof(AACEncContext),
.init = aac_encode_init,
2012-01-28 18:28:01 +01:00
.encode2 = aac_encode_frame,
.close = aac_encode_end,
.supported_samplerates = mpeg4audio_sample_rates,
.capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY |
AV_CODEC_CAP_EXPERIMENTAL,
2012-08-25 19:04:33 +02:00
.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
AV_SAMPLE_FMT_NONE },
.priv_class = &aacenc_class,
};