git.sesse.net Git - ffmpeg/blob - libavcodec/aacenc_tns.c

   1 /*
   2  * AAC encoder TNS
   3  * Copyright (C) 2015 Rostislav Pehlivanov
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23  * @file
  24  * AAC encoder temporal noise shaping
  25  * @author Rostislav Pehlivanov ( atomnuker gmail com )
  26  */
  27
  28 #include "aacenc.h"
  29 #include "aacenc_tns.h"
  30 #include "aactab.h"
  31 #include "aacenc_utils.h"
  32 #include "aacenc_quantization.h"
  33
  34 /**
  35  * Encode TNS data.
  36  * Coefficient compression saves a single bit per coefficient.
  37  */
  38 void ff_aac_encode_tns_info(AACEncContext *s, SingleChannelElement *sce)
  39 {
  40     uint8_t u_coef;
  41     const uint8_t coef_res = TNS_Q_BITS == 4;
  42     int i, w, filt, coef_len, coef_compress = 0;
  43     const int is8 = sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE;
  44     TemporalNoiseShaping *tns = &sce->tns;
  45
  46     if (!sce->tns.present)
  47         return;
  48
  49     for (i = 0; i < sce->ics.num_windows; i++) {
  50         put_bits(&s->pb, 2 - is8, sce->tns.n_filt[i]);
  51         if (tns->n_filt[i]) {
  52             put_bits(&s->pb, 1, coef_res);
  53             for (filt = 0; filt < tns->n_filt[i]; filt++) {
  54                 put_bits(&s->pb, 6 - 2 * is8, tns->length[i][filt]);
  55                 put_bits(&s->pb, 5 - 2 * is8, tns->order[i][filt]);
  56                 if (tns->order[i][filt]) {
  57                     put_bits(&s->pb, 1, !!tns->direction[i][filt]);
  58                     put_bits(&s->pb, 1, !!coef_compress);
  59                     coef_len = coef_res + 3 - coef_compress;
  60                     for (w = 0; w < tns->order[i][filt]; w++) {
  61                         u_coef = (tns->coef_idx[i][filt][w])&(~(~0<<coef_len));
  62                         put_bits(&s->pb, coef_len, u_coef);
  63                     }
  64                 }
  65             }
  66         }
  67     }
  68 }
  69
  70 static inline void quantize_coefs(double *coef, int *idx, float *lpc, int order)
  71 {
  72     int i;
  73     uint8_t u_coef;
  74     const float *quant_arr = tns_tmp2_map[TNS_Q_BITS == 4];
  75     const double iqfac_p = ((1 << (TNS_Q_BITS-1)) - 0.5)/(M_PI/2.0);
  76     const double iqfac_m = ((1 << (TNS_Q_BITS-1)) + 0.5)/(M_PI/2.0);
  77     for (i = 0; i < order; i++) {
  78         idx[i] = ceilf(asin(coef[i])*((coef[i] >= 0) ? iqfac_p : iqfac_m));
  79         u_coef = (idx[i])&(~(~0<<TNS_Q_BITS));
  80         lpc[i] = quant_arr[u_coef];
  81     }
  82 }
  83
  84 /* Apply TNS filter */
  85 void ff_aac_apply_tns(AACEncContext *s, SingleChannelElement *sce)
  86 {
  87     TemporalNoiseShaping *tns = &sce->tns;
  88     IndividualChannelStream *ics = &sce->ics;
  89     int w, filt, m, i, top, order, bottom, start, end, size, inc;
  90     const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
  91     float lpc[TNS_MAX_ORDER];
  92
  93     for (w = 0; w < ics->num_windows; w++) {
  94         bottom = ics->num_swb;
  95         for (filt = 0; filt < tns->n_filt[w]; filt++) {
  96             top    = bottom;
  97             bottom = FFMAX(0, top - tns->length[w][filt]);
  98             order  = tns->order[w][filt];
  99             if (order == 0)
 100                 continue;
 101
 102             // tns_decode_coef
 103             compute_lpc_coefs(tns->coef[w][filt], order, lpc, 0, 0, 0);
 104
 105             start = ics->swb_offset[FFMIN(bottom, mmm)];
 106             end   = ics->swb_offset[FFMIN(   top, mmm)];
 107             if ((size = end - start) <= 0)
 108                 continue;
 109             if (tns->direction[w][filt]) {
 110                 inc = -1;
 111                 start = end - 1;
 112             } else {
 113                 inc = 1;
 114             }
 115             start += w * 128;
 116
 117             // ar filter
 118             for (m = 0; m < size; m++, start += inc)
 119                 for (i = 1; i <= FFMIN(m, order); i++)
 120                     sce->coeffs[start] += lpc[i-1]*sce->pcoeffs[start - i*inc];
 121         }
 122     }
 123 }
 124
 125 void ff_aac_search_for_tns(AACEncContext *s, SingleChannelElement *sce)
 126 {
 127     TemporalNoiseShaping *tns = &sce->tns;
 128     int w, w2, g, count = 0;
 129     const int mmm = FFMIN(sce->ics.tns_max_bands, sce->ics.max_sfb);
 130     const int is8 = sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE;
 131     const int order = is8 ? 7 : s->profile == FF_PROFILE_AAC_LOW ? 12 : TNS_MAX_ORDER;
 132
 133     int sfb_start = av_clip(tns_min_sfb[is8][s->samplerate_index], 0, mmm);
 134     int sfb_end   = av_clip(sce->ics.num_swb, 0, mmm);
 135
 136     for (w = 0; w < sce->ics.num_windows; w++) {
 137         float e_ratio = 0.0f, threshold = 0.0f, spread = 0.0f, en[2] = {0.0, 0.0f};
 138         double gain = 0.0f, coefs[MAX_LPC_ORDER] = {0};
 139         int coef_start = w*sce->ics.num_swb + sce->ics.swb_offset[sfb_start];
 140         int coef_len = sce->ics.swb_offset[sfb_end] - sce->ics.swb_offset[sfb_start];
 141
 142         for (g = 0;  g < sce->ics.num_swb; g++) {
 143             if (w*16+g < sfb_start || w*16+g > sfb_end)
 144                 continue;
 145             for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
 146                 FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
 147                 if ((w+w2)*16+g > sfb_start + ((sfb_end - sfb_start)/2))
 148                     en[1] += band->energy;
 149                 else
 150                     en[0] += band->energy;
 151                 threshold += band->threshold;
 152                 spread += band->spread;
 153             }
 154         }
 155
 156         if (coef_len <= 0 || (sfb_end - sfb_start) <= 0)
 157             continue;
 158         else
 159             e_ratio = en[0]/en[1];
 160
 161         /* LPC */
 162         gain = ff_lpc_calc_ref_coefs_f(&s->lpc, &sce->coeffs[coef_start],
 163                                        coef_len, order, coefs);
 164
 165         if (gain > TNS_GAIN_THRESHOLD_LOW && gain < TNS_GAIN_THRESHOLD_HIGH &&
 166             (en[0]+en[1]) > TNS_GAIN_THRESHOLD_LOW*threshold &&
 167             spread < TNS_SPREAD_THRESHOLD && order) {
 168             if (is8 || order < 2 || (e_ratio > TNS_E_RATIO_LOW && e_ratio < TNS_E_RATIO_HIGH)) {
 169                 tns->n_filt[w] = 1;
 170                 for (g = 0; g < tns->n_filt[w]; g++) {
 171                     tns->length[w][g] = sfb_end - sfb_start;
 172                     tns->direction[w][g] = en[0] < en[1];
 173                     tns->order[w][g] = order;
 174                     quantize_coefs(coefs, tns->coef_idx[w][g], tns->coef[w][g],
 175                                    order);
 176                 }
 177             } else {  /* 2 filters due to energy disbalance */
 178                 tns->n_filt[w] = 2;
 179                 for (g = 0; g < tns->n_filt[w]; g++) {
 180                     tns->direction[w][g] = en[g] < en[!g];
 181                     tns->order[w][g] = !g ? order/2 : order - tns->order[w][g-1];
 182                     tns->length[w][g] = !g ? (sfb_end - sfb_start)/2 : \
 183                                     (sfb_end - sfb_start) - tns->length[w][g-1];
 184                     quantize_coefs(&coefs[!g ? 0 : order - tns->order[w][g-1]],
 185                                    tns->coef_idx[w][g], tns->coef[w][g],
 186                                    tns->order[w][g]);
 187                 }
 188             }
 189             count++;
 190         }
 191     }
 192
 193     sce->tns.present = !!count;
 194 }