| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * AAC coefficients encoder | ||
| 3 | * Copyright (C) 2008-2009 Konstantin Shishkov | ||
| 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 coefficients encoder | ||
| 25 | */ | ||
| 26 | |||
| 27 | /*********************************** | ||
| 28 | * TODOs: | ||
| 29 | * speedup quantizer selection | ||
| 30 | * add sane pulse detection | ||
| 31 | ***********************************/ | ||
| 32 | |||
| 33 | #include <math.h> // brought forward to work around cygwin header breakage | ||
| 34 | |||
| 35 | #include <float.h> | ||
| 36 | |||
| 37 | #include "libavutil/mathematics.h" | ||
| 38 | #include "mathops.h" | ||
| 39 | #include "avcodec.h" | ||
| 40 | #include "put_bits.h" | ||
| 41 | #include "aac.h" | ||
| 42 | #include "aacenc.h" | ||
| 43 | #include "aactab.h" | ||
| 44 | #include "aacenctab.h" | ||
| 45 | #include "aacenc_utils.h" | ||
| 46 | #include "aacenc_quantization.h" | ||
| 47 | |||
| 48 | #include "aacenc_is.h" | ||
| 49 | #include "aacenc_tns.h" | ||
| 50 | |||
| 51 | #include "libavcodec/aaccoder_twoloop.h" | ||
| 52 | |||
| 53 | /* Parameter of f(x) = a*(lambda/100), defines the maximum fourier spread | ||
| 54 | * beyond which no PNS is used (since the SFBs contain tone rather than noise) */ | ||
| 55 | #define NOISE_SPREAD_THRESHOLD 0.9f | ||
| 56 | |||
| 57 | /* Parameter of f(x) = a*(100/lambda), defines how much PNS is allowed to | ||
| 58 | * replace low energy non zero bands */ | ||
| 59 | #define NOISE_LAMBDA_REPLACE 1.948f | ||
| 60 | |||
| 61 | #include "libavcodec/aaccoder_trellis.h" | ||
| 62 | #include "libavcodec/aaccoder_nmr.h" | ||
| 63 | |||
| 64 | typedef float (*quantize_and_encode_band_func)(struct AACEncContext *s, PutBitContext *pb, | ||
| 65 | const float *in, float *quant, const float *scaled, | ||
| 66 | int size, int scale_idx, int cb, | ||
| 67 | const float lambda, const float uplim, | ||
| 68 | int *bits, float *energy); | ||
| 69 | |||
| 70 | /** | ||
| 71 | * Calculate rate distortion cost for quantizing with given codebook | ||
| 72 | * | ||
| 73 | * @return quantization distortion | ||
| 74 | */ | ||
| 75 | 11763085 | static av_always_inline float quantize_and_encode_band_cost_template( | |
| 76 | struct AACEncContext *s, | ||
| 77 | PutBitContext *pb, const float *in, float *out, | ||
| 78 | const float *scaled, int size, int scale_idx, | ||
| 79 | int cb, const float lambda, const float uplim, | ||
| 80 | int *bits, float *energy, int BT_ZERO, int BT_UNSIGNED, | ||
| 81 | int BT_PAIR, int BT_ESC, int BT_NOISE, int BT_STEREO, | ||
| 82 | const float ROUNDING) | ||
| 83 | { | ||
| 84 | 11763085 | const int q_idx = POW_SF2_ZERO - scale_idx + SCALE_ONE_POS - SCALE_DIV_512; | |
| 85 | 11763085 | const float Q = ff_aac_pow2sf_tab [q_idx]; | |
| 86 | 11763085 | const float Q34 = ff_aac_pow34sf_tab[q_idx]; | |
| 87 | 11763085 | const float IQ = ff_aac_pow2sf_tab [POW_SF2_ZERO + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; | |
| 88 | 11763085 | const float CLIPPED_ESCAPE = 165140.0f*IQ; | |
| 89 | 11763085 | float cost = 0; | |
| 90 | 11763085 | float qenergy = 0; | |
| 91 |
2/2✓ Branch 0 taken 6798084 times.
✓ Branch 1 taken 4965001 times.
|
11763085 | const int dim = BT_PAIR ? 2 : 4; |
| 92 | 11763085 | int resbits = 0; | |
| 93 | int off; | ||
| 94 | |||
| 95 |
6/6✓ Branch 0 taken 10585936 times.
✓ Branch 1 taken 1177149 times.
✓ Branch 2 taken 10578986 times.
✓ Branch 3 taken 6950 times.
✓ Branch 4 taken 47228 times.
✓ Branch 5 taken 10531758 times.
|
11763085 | if (BT_ZERO || BT_NOISE || BT_STEREO) { |
| 96 |
2/2✓ Branch 0 taken 25646100 times.
✓ Branch 1 taken 1231327 times.
|
26877427 | for (int i = 0; i < size; i++) |
| 97 | 25646100 | cost += in[i]*in[i]; | |
| 98 |
2/2✓ Branch 0 taken 1198901 times.
✓ Branch 1 taken 32426 times.
|
1231327 | if (bits) |
| 99 | 1198901 | *bits = 0; | |
| 100 |
2/2✓ Branch 0 taken 1159697 times.
✓ Branch 1 taken 71630 times.
|
1231327 | if (energy) |
| 101 | 1159697 | *energy = qenergy; | |
| 102 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1231327 times.
|
1231327 | if (out) { |
| 103 | ✗ | for (int i = 0; i < size; i += dim) | |
| 104 | ✗ | for (int j = 0; j < dim; j++) | |
| 105 | ✗ | out[i+j] = 0.0f; | |
| 106 | } | ||
| 107 | 1231327 | return cost * lambda; | |
| 108 | } | ||
| 109 |
2/2✓ Branch 0 taken 519170 times.
✓ Branch 1 taken 10012588 times.
|
10531758 | if (!scaled) { |
| 110 | 519170 | s->aacdsp.abs_pow34(s->scoefs, in, size); | |
| 111 | 519170 | scaled = s->scoefs; | |
| 112 | } | ||
| 113 | 10531758 | s->aacdsp.quant_bands(s->qcoefs, in, scaled, size, !BT_UNSIGNED, aac_cb_maxval[cb], Q34, ROUNDING); | |
| 114 |
2/2✓ Branch 0 taken 6864499 times.
✓ Branch 1 taken 3667259 times.
|
10531758 | if (BT_UNSIGNED) { |
| 115 | 6864499 | off = 0; | |
| 116 | } else { | ||
| 117 | 3667259 | off = aac_cb_maxval[cb]; | |
| 118 | } | ||
| 119 |
2/2✓ Branch 0 taken 77476221 times.
✓ Branch 1 taken 10531758 times.
|
88007979 | for (int i = 0; i < size; i += dim) { |
| 120 | const float *vec; | ||
| 121 | 77476221 | int *quants = s->qcoefs + i; | |
| 122 | 77476221 | int curidx = 0; | |
| 123 | int curbits; | ||
| 124 | 77476221 | float quantized, rd = 0.0f; | |
| 125 |
2/2✓ Branch 0 taken 193891332 times.
✓ Branch 1 taken 77476221 times.
|
271367553 | for (int j = 0; j < dim; j++) { |
| 126 | 193891332 | curidx *= aac_cb_range[cb]; | |
| 127 | 193891332 | curidx += quants[j] + off; | |
| 128 | } | ||
| 129 | 77476221 | curbits = ff_aac_spectral_bits[cb-1][curidx]; | |
| 130 | 77476221 | vec = &ff_aac_codebook_vectors[cb-1][curidx*dim]; | |
| 131 |
2/2✓ Branch 0 taken 51790259 times.
✓ Branch 1 taken 25685962 times.
|
77476221 | if (BT_UNSIGNED) { |
| 132 |
2/2✓ Branch 0 taken 120077444 times.
✓ Branch 1 taken 51790259 times.
|
171867703 | for (int j = 0; j < dim; j++) { |
| 133 | 120077444 | float t = fabsf(in[i+j]); | |
| 134 | float di; | ||
| 135 |
4/4✓ Branch 0 taken 41286660 times.
✓ Branch 1 taken 78790784 times.
✓ Branch 2 taken 20113827 times.
✓ Branch 3 taken 21172833 times.
|
120077444 | if (BT_ESC && vec[j] == 64.0f) { //FIXME: slow |
| 136 |
2/2✓ Branch 0 taken 2092313 times.
✓ Branch 1 taken 18021514 times.
|
20113827 | if (t >= CLIPPED_ESCAPE) { |
| 137 | 2092313 | quantized = CLIPPED_ESCAPE; | |
| 138 | 2092313 | curbits += 21; | |
| 139 | } else { | ||
| 140 | 18021514 | int c = av_clip_uintp2(quant(t, Q, ROUNDING), 13); | |
| 141 | 18021514 | quantized = c*cbrtf(c)*IQ; | |
| 142 | 18021514 | curbits += av_log2(c)*2 - 4 + 1; | |
| 143 | } | ||
| 144 | } else { | ||
| 145 | 99963617 | quantized = vec[j]*IQ; | |
| 146 | } | ||
| 147 | 120077444 | di = t - quantized; | |
| 148 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 120077444 times.
|
120077444 | if (out) |
| 149 | ✗ | out[i+j] = in[i+j] >= 0 ? quantized : -quantized; | |
| 150 |
2/2✓ Branch 0 taken 81753851 times.
✓ Branch 1 taken 38323593 times.
|
120077444 | if (vec[j] != 0.0f) |
| 151 | 81753851 | curbits++; | |
| 152 | 120077444 | qenergy += quantized*quantized; | |
| 153 | 120077444 | rd += di*di; | |
| 154 | } | ||
| 155 | } else { | ||
| 156 |
2/2✓ Branch 0 taken 73813888 times.
✓ Branch 1 taken 25685962 times.
|
99499850 | for (int j = 0; j < dim; j++) { |
| 157 | 73813888 | quantized = vec[j]*IQ; | |
| 158 | 73813888 | qenergy += quantized*quantized; | |
| 159 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 73813888 times.
|
73813888 | if (out) |
| 160 | ✗ | out[i+j] = quantized; | |
| 161 | 73813888 | rd += (in[i+j] - quantized)*(in[i+j] - quantized); | |
| 162 | } | ||
| 163 | } | ||
| 164 | 77476221 | cost += rd * lambda + curbits; | |
| 165 | 77476221 | resbits += curbits; | |
| 166 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 77476221 times.
|
77476221 | if (cost >= uplim) |
| 167 | ✗ | return uplim; | |
| 168 |
2/2✓ Branch 0 taken 3521406 times.
✓ Branch 1 taken 73954815 times.
|
77476221 | if (pb) { |
| 169 | 3521406 | put_bits(pb, ff_aac_spectral_bits[cb-1][curidx], ff_aac_spectral_codes[cb-1][curidx]); | |
| 170 |
2/2✓ Branch 0 taken 1781153 times.
✓ Branch 1 taken 1740253 times.
|
3521406 | if (BT_UNSIGNED) |
| 171 |
2/2✓ Branch 0 taken 4875576 times.
✓ Branch 1 taken 1781153 times.
|
6656729 | for (int j = 0; j < dim; j++) |
| 172 |
2/2✓ Branch 0 taken 3336259 times.
✓ Branch 1 taken 1539317 times.
|
4875576 | if (ff_aac_codebook_vectors[cb-1][curidx*dim+j] != 0.0f) |
| 173 | 3336259 | put_bits(pb, 1, in[i+j] < 0.0f); | |
| 174 |
2/2✓ Branch 0 taken 681990 times.
✓ Branch 1 taken 2839416 times.
|
3521406 | if (BT_ESC) { |
| 175 |
2/2✓ Branch 0 taken 1363980 times.
✓ Branch 1 taken 681990 times.
|
2045970 | for (int j = 0; j < 2; j++) { |
| 176 |
2/2✓ Branch 0 taken 196902 times.
✓ Branch 1 taken 1167078 times.
|
1363980 | if (ff_aac_codebook_vectors[cb-1][curidx*2+j] == 64.0f) { |
| 177 | 196902 | int coef = av_clip(quant(fabsf(in[i+j]), Q, ROUNDING), 16, (1 << 13) - 1); | |
| 178 | 196902 | int len = av_log2(coef); | |
| 179 | |||
| 180 | 196902 | put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2); | |
| 181 | 196902 | put_sbits(pb, len, coef); | |
| 182 | } | ||
| 183 | } | ||
| 184 | } | ||
| 185 | } | ||
| 186 | } | ||
| 187 | |||
| 188 |
2/2✓ Branch 0 taken 9944537 times.
✓ Branch 1 taken 587221 times.
|
10531758 | if (bits) |
| 189 | 9944537 | *bits = resbits; | |
| 190 |
2/2✓ Branch 0 taken 6003127 times.
✓ Branch 1 taken 4528631 times.
|
10531758 | if (energy) |
| 191 | 6003127 | *energy = qenergy; | |
| 192 | 10531758 | return cost; | |
| 193 | } | ||
| 194 | |||
| 195 | ✗ | static inline float quantize_and_encode_band_cost_NONE(struct AACEncContext *s, PutBitContext *pb, | |
| 196 | const float *in, float *quant, const float *scaled, | ||
| 197 | int size, int scale_idx, int cb, | ||
| 198 | const float lambda, const float uplim, | ||
| 199 | int *bits, float *energy) { | ||
| 200 | ✗ | av_assert0(0); | |
| 201 | return 0.0f; | ||
| 202 | } | ||
| 203 | |||
| 204 | #define QUANTIZE_AND_ENCODE_BAND_COST_FUNC(NAME, BT_ZERO, BT_UNSIGNED, BT_PAIR, BT_ESC, BT_NOISE, BT_STEREO, ROUNDING) \ | ||
| 205 | static float quantize_and_encode_band_cost_ ## NAME( \ | ||
| 206 | struct AACEncContext *s, \ | ||
| 207 | PutBitContext *pb, const float *in, float *quant, \ | ||
| 208 | const float *scaled, int size, int scale_idx, \ | ||
| 209 | int cb, const float lambda, const float uplim, \ | ||
| 210 | int *bits, float *energy) { \ | ||
| 211 | return quantize_and_encode_band_cost_template( \ | ||
| 212 | s, pb, in, quant, scaled, size, scale_idx, \ | ||
| 213 | BT_ESC ? ESC_BT : cb, lambda, uplim, bits, energy, \ | ||
| 214 | BT_ZERO, BT_UNSIGNED, BT_PAIR, BT_ESC, BT_NOISE, BT_STEREO, \ | ||
| 215 | ROUNDING); \ | ||
| 216 | } | ||
| 217 | |||
| 218 | 1177149 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ZERO, 1, 0, 0, 0, 0, 0, ROUND_STANDARD) | |
| 219 | 2146224 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(SQUAD, 0, 0, 0, 0, 0, 0, ROUND_STANDARD) | |
| 220 | 1587450 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(UQUAD, 0, 1, 0, 0, 0, 0, ROUND_STANDARD) | |
| 221 | 1521035 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(SPAIR, 0, 0, 1, 0, 0, 0, ROUND_STANDARD) | |
| 222 | 2513870 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(UPAIR, 0, 1, 1, 0, 0, 0, ROUND_STANDARD) | |
| 223 | 2760691 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ESC, 0, 1, 1, 1, 0, 0, ROUND_STANDARD) | |
| 224 | 2488 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ESC_RTZ, 0, 1, 1, 1, 0, 0, ROUND_TO_ZERO) | |
| 225 | 6950 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(NOISE, 0, 0, 0, 0, 1, 0, ROUND_STANDARD) | |
| 226 | 47228 | QUANTIZE_AND_ENCODE_BAND_COST_FUNC(STEREO,0, 0, 0, 0, 0, 1, ROUND_STANDARD) | |
| 227 | |||
| 228 | static const quantize_and_encode_band_func quantize_and_encode_band_cost_arr[] = | ||
| 229 | { | ||
| 230 | quantize_and_encode_band_cost_ZERO, | ||
| 231 | quantize_and_encode_band_cost_SQUAD, | ||
| 232 | quantize_and_encode_band_cost_SQUAD, | ||
| 233 | quantize_and_encode_band_cost_UQUAD, | ||
| 234 | quantize_and_encode_band_cost_UQUAD, | ||
| 235 | quantize_and_encode_band_cost_SPAIR, | ||
| 236 | quantize_and_encode_band_cost_SPAIR, | ||
| 237 | quantize_and_encode_band_cost_UPAIR, | ||
| 238 | quantize_and_encode_band_cost_UPAIR, | ||
| 239 | quantize_and_encode_band_cost_UPAIR, | ||
| 240 | quantize_and_encode_band_cost_UPAIR, | ||
| 241 | quantize_and_encode_band_cost_ESC, | ||
| 242 | quantize_and_encode_band_cost_NONE, /* CB 12 doesn't exist */ | ||
| 243 | quantize_and_encode_band_cost_NOISE, | ||
| 244 | quantize_and_encode_band_cost_STEREO, | ||
| 245 | quantize_and_encode_band_cost_STEREO, | ||
| 246 | }; | ||
| 247 | |||
| 248 | static const quantize_and_encode_band_func quantize_and_encode_band_cost_rtz_arr[] = | ||
| 249 | { | ||
| 250 | quantize_and_encode_band_cost_ZERO, | ||
| 251 | quantize_and_encode_band_cost_SQUAD, | ||
| 252 | quantize_and_encode_band_cost_SQUAD, | ||
| 253 | quantize_and_encode_band_cost_UQUAD, | ||
| 254 | quantize_and_encode_band_cost_UQUAD, | ||
| 255 | quantize_and_encode_band_cost_SPAIR, | ||
| 256 | quantize_and_encode_band_cost_SPAIR, | ||
| 257 | quantize_and_encode_band_cost_UPAIR, | ||
| 258 | quantize_and_encode_band_cost_UPAIR, | ||
| 259 | quantize_and_encode_band_cost_UPAIR, | ||
| 260 | quantize_and_encode_band_cost_UPAIR, | ||
| 261 | quantize_and_encode_band_cost_ESC_RTZ, | ||
| 262 | quantize_and_encode_band_cost_NONE, /* CB 12 doesn't exist */ | ||
| 263 | quantize_and_encode_band_cost_NOISE, | ||
| 264 | quantize_and_encode_band_cost_STEREO, | ||
| 265 | quantize_and_encode_band_cost_STEREO, | ||
| 266 | }; | ||
| 267 | |||
| 268 | 11216826 | float ff_quantize_and_encode_band_cost(struct AACEncContext *s, PutBitContext *pb, | |
| 269 | const float *in, float *quant, const float *scaled, | ||
| 270 | int size, int scale_idx, int cb, | ||
| 271 | const float lambda, const float uplim, | ||
| 272 | int *bits, float *energy) | ||
| 273 | { | ||
| 274 | 11216826 | return quantize_and_encode_band_cost_arr[cb](s, pb, in, quant, scaled, size, | |
| 275 | scale_idx, cb, lambda, uplim, | ||
| 276 | bits, energy); | ||
| 277 | } | ||
| 278 | |||
| 279 | 546259 | static inline void quantize_and_encode_band(struct AACEncContext *s, PutBitContext *pb, | |
| 280 | const float *in, float *out, int size, int scale_idx, | ||
| 281 | int cb, const float lambda, int rtz) | ||
| 282 | { | ||
| 283 |
2/2✓ Branch 0 taken 5350 times.
✓ Branch 1 taken 540909 times.
|
546259 | (rtz ? quantize_and_encode_band_cost_rtz_arr : quantize_and_encode_band_cost_arr)[cb](s, pb, in, out, NULL, size, scale_idx, cb, |
| 284 | lambda, INFINITY, NULL, NULL); | ||
| 285 | 546259 | } | |
| 286 | |||
| 287 | /** | ||
| 288 | * structure used in optimal codebook search | ||
| 289 | */ | ||
| 290 | typedef struct BandCodingPath { | ||
| 291 | int prev_idx; ///< pointer to the previous path point | ||
| 292 | float cost; ///< path cost | ||
| 293 | int run; | ||
| 294 | } BandCodingPath; | ||
| 295 | |||
| 296 | typedef struct TrellisPath { | ||
| 297 | float cost; | ||
| 298 | int prev; | ||
| 299 | } TrellisPath; | ||
| 300 | |||
| 301 | #define TRELLIS_STAGES 121 | ||
| 302 | #define TRELLIS_STATES (SCALE_MAX_DIFF+1) | ||
| 303 | |||
| 304 | 11374 | static void set_special_band_scalefactors(AACEncContext *s, SingleChannelElement *sce) | |
| 305 | { | ||
| 306 | int w, g; | ||
| 307 | 11374 | int prevscaler_n = -255, prevscaler_i = 0; | |
| 308 | 11374 | int bands = 0; | |
| 309 | |||
| 310 |
2/2✓ Branch 0 taken 12089 times.
✓ Branch 1 taken 11374 times.
|
23463 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 311 |
2/2✓ Branch 0 taken 541693 times.
✓ Branch 1 taken 12089 times.
|
553782 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 312 |
2/2✓ Branch 0 taken 23663 times.
✓ Branch 1 taken 518030 times.
|
541693 | if (sce->zeroes[w*16+g]) |
| 313 | 23663 | continue; | |
| 314 |
4/4✓ Branch 0 taken 497634 times.
✓ Branch 1 taken 20396 times.
✓ Branch 2 taken 2390 times.
✓ Branch 3 taken 495244 times.
|
518030 | if (sce->band_type[w*16+g] == INTENSITY_BT || sce->band_type[w*16+g] == INTENSITY_BT2) { |
| 315 | 22786 | sce->sf_idx[w*16+g] = av_clip(roundf(log2f(sce->is_ener[w*16+g])*2), -155, 100); | |
| 316 | 22786 | bands++; | |
| 317 |
2/2✓ Branch 0 taken 3416 times.
✓ Branch 1 taken 491828 times.
|
495244 | } else if (sce->band_type[w*16+g] == NOISE_BT) { |
| 318 | 3416 | sce->sf_idx[w*16+g] = av_clip(3+ceilf(log2f(sce->pns_ener[w*16+g])*2), -100, 155); | |
| 319 |
2/2✓ Branch 0 taken 842 times.
✓ Branch 1 taken 2574 times.
|
3416 | if (prevscaler_n == -255) |
| 320 | 842 | prevscaler_n = sce->sf_idx[w*16+g]; | |
| 321 | 3416 | bands++; | |
| 322 | } | ||
| 323 | } | ||
| 324 | } | ||
| 325 | |||
| 326 |
2/2✓ Branch 0 taken 9033 times.
✓ Branch 1 taken 2341 times.
|
11374 | if (!bands) |
| 327 | 9033 | return; | |
| 328 | |||
| 329 | /* Clip the scalefactor indices */ | ||
| 330 |
2/2✓ Branch 0 taken 2414 times.
✓ Branch 1 taken 2341 times.
|
4755 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 331 |
2/2✓ Branch 0 taken 114156 times.
✓ Branch 1 taken 2414 times.
|
116570 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 332 |
2/2✓ Branch 0 taken 6892 times.
✓ Branch 1 taken 107264 times.
|
114156 | if (sce->zeroes[w*16+g]) |
| 333 | 6892 | continue; | |
| 334 |
4/4✓ Branch 0 taken 86868 times.
✓ Branch 1 taken 20396 times.
✓ Branch 2 taken 2390 times.
✓ Branch 3 taken 84478 times.
|
107264 | if (sce->band_type[w*16+g] == INTENSITY_BT || sce->band_type[w*16+g] == INTENSITY_BT2) { |
| 335 | 22786 | sce->sf_idx[w*16+g] = prevscaler_i = av_clip(sce->sf_idx[w*16+g], prevscaler_i - SCALE_MAX_DIFF, prevscaler_i + SCALE_MAX_DIFF); | |
| 336 |
2/2✓ Branch 0 taken 3416 times.
✓ Branch 1 taken 81062 times.
|
84478 | } else if (sce->band_type[w*16+g] == NOISE_BT) { |
| 337 | 3416 | sce->sf_idx[w*16+g] = prevscaler_n = av_clip(sce->sf_idx[w*16+g], prevscaler_n - SCALE_MAX_DIFF, prevscaler_n + SCALE_MAX_DIFF); | |
| 338 | } | ||
| 339 | } | ||
| 340 | } | ||
| 341 | } | ||
| 342 | |||
| 343 | 7966 | static void search_for_quantizers_fast(AVCodecContext *avctx, AACEncContext *s, | |
| 344 | SingleChannelElement *sce, | ||
| 345 | const float lambda) | ||
| 346 | { | ||
| 347 | 7966 | int start = 0, i, w, w2, g; | |
| 348 | 7966 | int destbits = avctx->bit_rate * 1024.0 / avctx->sample_rate / avctx->ch_layout.nb_channels * (lambda / 120.f); | |
| 349 | 7966 | float dists[128] = { 0 }, uplims[128] = { 0 }; | |
| 350 | float maxvals[128]; | ||
| 351 | int fflag, minscaler; | ||
| 352 | 7966 | int its = 0; | |
| 353 | 7966 | int allz = 0; | |
| 354 | 7966 | float minthr = INFINITY; | |
| 355 | |||
| 356 | // for values above this the decoder might end up in an endless loop | ||
| 357 | // due to always having more bits than what can be encoded. | ||
| 358 | 7966 | destbits = FFMIN(destbits, 5800); | |
| 359 | //some heuristic to determine initial quantizers will reduce search time | ||
| 360 | //determine zero bands and upper limits | ||
| 361 |
2/2✓ Branch 0 taken 8418 times.
✓ Branch 1 taken 7966 times.
|
16384 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 362 | 8418 | start = 0; | |
| 363 |
2/2✓ Branch 0 taken 382684 times.
✓ Branch 1 taken 8418 times.
|
391102 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 364 | 382684 | int nz = 0; | |
| 365 | 382684 | float uplim = 0.0f; | |
| 366 |
2/2✓ Branch 0 taken 406736 times.
✓ Branch 1 taken 382684 times.
|
789420 | for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) { |
| 367 | 406736 | FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g]; | |
| 368 | 406736 | uplim += band->threshold; | |
| 369 |
3/4✓ Branch 0 taken 403343 times.
✓ Branch 1 taken 3393 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 403343 times.
|
406736 | if (band->energy <= band->threshold || band->threshold == 0.0f) { |
| 370 | 3393 | sce->zeroes[(w+w2)*16+g] = 1; | |
| 371 | 3393 | continue; | |
| 372 | } | ||
| 373 | 403343 | nz = 1; | |
| 374 | } | ||
| 375 | 382684 | uplims[w*16+g] = uplim *512; | |
| 376 | 382684 | sce->band_type[w*16+g] = 0; | |
| 377 | 382684 | sce->zeroes[w*16+g] = !nz; | |
| 378 |
2/2✓ Branch 0 taken 379311 times.
✓ Branch 1 taken 3373 times.
|
382684 | if (nz) |
| 379 |
2/2✓ Branch 0 taken 94427 times.
✓ Branch 1 taken 284884 times.
|
379311 | minthr = FFMIN(minthr, uplim); |
| 380 | 382684 | allz |= nz; | |
| 381 | 382684 | start += sce->ics.swb_sizes[g]; | |
| 382 | } | ||
| 383 | } | ||
| 384 |
2/2✓ Branch 0 taken 8418 times.
✓ Branch 1 taken 7966 times.
|
16384 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 385 |
2/2✓ Branch 0 taken 382684 times.
✓ Branch 1 taken 8418 times.
|
391102 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 386 |
2/2✓ Branch 0 taken 3373 times.
✓ Branch 1 taken 379311 times.
|
382684 | if (sce->zeroes[w*16+g]) { |
| 387 | 3373 | sce->sf_idx[w*16+g] = SCALE_ONE_POS; | |
| 388 | 3373 | continue; | |
| 389 | } | ||
| 390 |
2/2✓ Branch 0 taken 188225 times.
✓ Branch 1 taken 191086 times.
|
379311 | sce->sf_idx[w*16+g] = SCALE_ONE_POS + FFMIN(log2f(uplims[w*16+g]/minthr)*4,59); |
| 391 | } | ||
| 392 | } | ||
| 393 | |||
| 394 |
2/2✓ Branch 0 taken 12 times.
✓ Branch 1 taken 7954 times.
|
7966 | if (!allz) |
| 395 | 12 | return; | |
| 396 | 7954 | s->aacdsp.abs_pow34(s->scoefs, sce->coeffs, 1024); | |
| 397 | 7954 | ff_quantize_band_cost_cache_init(s); | |
| 398 | |||
| 399 |
2/2✓ Branch 0 taken 8406 times.
✓ Branch 1 taken 7954 times.
|
16360 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 400 | 8406 | start = w*128; | |
| 401 |
2/2✓ Branch 0 taken 382096 times.
✓ Branch 1 taken 8406 times.
|
390502 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 402 | 382096 | const float *scaled = s->scoefs + start; | |
| 403 | 382096 | maxvals[w*16+g] = find_max_val(sce->ics.group_len[w], sce->ics.swb_sizes[g], scaled); | |
| 404 | 382096 | start += sce->ics.swb_sizes[g]; | |
| 405 | } | ||
| 406 | } | ||
| 407 | |||
| 408 | //perform two-loop search | ||
| 409 | //outer loop - improve quality | ||
| 410 | do { | ||
| 411 | int tbits, qstep; | ||
| 412 | 34641 | minscaler = sce->sf_idx[0]; | |
| 413 | //inner loop - quantize spectrum to fit into given number of bits | ||
| 414 |
2/2✓ Branch 0 taken 26687 times.
✓ Branch 1 taken 7954 times.
|
34641 | qstep = its ? 1 : 32; |
| 415 | do { | ||
| 416 | 78062 | int prev = -1; | |
| 417 | 78062 | tbits = 0; | |
| 418 |
2/2✓ Branch 0 taken 82834 times.
✓ Branch 1 taken 78062 times.
|
160896 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 419 | 82834 | start = w*128; | |
| 420 |
2/2✓ Branch 0 taken 3733941 times.
✓ Branch 1 taken 82834 times.
|
3816775 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 421 | 3733941 | const float *coefs = sce->coeffs + start; | |
| 422 | 3733941 | const float *scaled = s->scoefs + start; | |
| 423 | 3733941 | int bits = 0; | |
| 424 | int cb; | ||
| 425 | 3733941 | float dist = 0.0f; | |
| 426 | |||
| 427 |
3/4✓ Branch 0 taken 3693559 times.
✓ Branch 1 taken 40382 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 3693559 times.
|
3733941 | if (sce->zeroes[w*16+g] || sce->sf_idx[w*16+g] >= 218) { |
| 428 | 40382 | start += sce->ics.swb_sizes[g]; | |
| 429 | 40382 | continue; | |
| 430 | } | ||
| 431 | 3693559 | minscaler = FFMIN(minscaler, sce->sf_idx[w*16+g]); | |
| 432 | 3693559 | cb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]); | |
| 433 |
2/2✓ Branch 0 taken 3937209 times.
✓ Branch 1 taken 3693559 times.
|
7630768 | for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) { |
| 434 | int b; | ||
| 435 | 7874418 | dist += quantize_band_cost_cached(s, w + w2, g, | |
| 436 | 3937209 | coefs + w2*128, | |
| 437 | 3937209 | scaled + w2*128, | |
| 438 | 3937209 | sce->ics.swb_sizes[g], | |
| 439 | 3937209 | sce->sf_idx[w*16+g], | |
| 440 | cb, 1.0f, INFINITY, | ||
| 441 | &b, NULL, 0); | ||
| 442 | 3937209 | bits += b; | |
| 443 | } | ||
| 444 | 3693559 | dists[w*16+g] = dist - bits; | |
| 445 |
2/2✓ Branch 0 taken 3615497 times.
✓ Branch 1 taken 78062 times.
|
3693559 | if (prev != -1) { |
| 446 | 3615497 | bits += ff_aac_scalefactor_bits[sce->sf_idx[w*16+g] - prev + SCALE_DIFF_ZERO]; | |
| 447 | } | ||
| 448 | 3693559 | tbits += bits; | |
| 449 | 3693559 | start += sce->ics.swb_sizes[g]; | |
| 450 | 3693559 | prev = sce->sf_idx[w*16+g]; | |
| 451 | } | ||
| 452 | } | ||
| 453 |
2/2✓ Branch 0 taken 14509 times.
✓ Branch 1 taken 63553 times.
|
78062 | if (tbits > destbits) { |
| 454 |
2/2✓ Branch 0 taken 1857152 times.
✓ Branch 1 taken 14509 times.
|
1871661 | for (i = 0; i < 128; i++) |
| 455 |
1/2✓ Branch 0 taken 1857152 times.
✗ Branch 1 not taken.
|
1857152 | if (sce->sf_idx[i] < 218 - qstep) |
| 456 | 1857152 | sce->sf_idx[i] += qstep; | |
| 457 | } else { | ||
| 458 |
2/2✓ Branch 0 taken 8134784 times.
✓ Branch 1 taken 63553 times.
|
8198337 | for (i = 0; i < 128; i++) |
| 459 |
2/2✓ Branch 0 taken 3283891 times.
✓ Branch 1 taken 4850893 times.
|
8134784 | if (sce->sf_idx[i] > 60 - qstep) |
| 460 | 3283891 | sce->sf_idx[i] -= qstep; | |
| 461 | } | ||
| 462 | 78062 | qstep >>= 1; | |
| 463 |
5/6✓ Branch 0 taken 38292 times.
✓ Branch 1 taken 39770 times.
✓ Branch 2 taken 3651 times.
✓ Branch 3 taken 34641 times.
✓ Branch 4 taken 3651 times.
✗ Branch 5 not taken.
|
78062 | if (!qstep && tbits > destbits*1.02 && sce->sf_idx[0] < 217) |
| 464 | 3651 | qstep = 1; | |
| 465 |
2/2✓ Branch 0 taken 43421 times.
✓ Branch 1 taken 34641 times.
|
78062 | } while (qstep); |
| 466 | |||
| 467 | 34641 | fflag = 0; | |
| 468 | 34641 | minscaler = av_clip(minscaler, 60, 255 - SCALE_MAX_DIFF); | |
| 469 | |||
| 470 |
2/2✓ Branch 0 taken 36113 times.
✓ Branch 1 taken 34641 times.
|
70754 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 471 |
2/2✓ Branch 0 taken 1653032 times.
✓ Branch 1 taken 36113 times.
|
1689145 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 472 | 1653032 | int prevsc = sce->sf_idx[w*16+g]; | |
| 473 |
3/4✓ Branch 0 taken 5417 times.
✓ Branch 1 taken 1647615 times.
✓ Branch 2 taken 5417 times.
✗ Branch 3 not taken.
|
1653032 | if (dists[w*16+g] > uplims[w*16+g] && sce->sf_idx[w*16+g] > 60) { |
| 474 |
2/2✓ Branch 1 taken 5239 times.
✓ Branch 2 taken 178 times.
|
5417 | if (find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]-1)) |
| 475 | 5239 | sce->sf_idx[w*16+g]--; | |
| 476 | else //Try to make sure there is some energy in every band | ||
| 477 | 178 | sce->sf_idx[w*16+g]-=2; | |
| 478 | } | ||
| 479 | 1653032 | sce->sf_idx[w*16+g] = av_clip(sce->sf_idx[w*16+g], minscaler, minscaler + SCALE_MAX_DIFF); | |
| 480 | 1653032 | sce->sf_idx[w*16+g] = FFMIN(sce->sf_idx[w*16+g], 219); | |
| 481 |
2/2✓ Branch 0 taken 388376 times.
✓ Branch 1 taken 1264656 times.
|
1653032 | if (sce->sf_idx[w*16+g] != prevsc) |
| 482 | 388376 | fflag = 1; | |
| 483 | 1653032 | sce->band_type[w*16+g] = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]); | |
| 484 | } | ||
| 485 | } | ||
| 486 | 34641 | its++; | |
| 487 |
4/4✓ Branch 0 taken 29596 times.
✓ Branch 1 taken 5045 times.
✓ Branch 2 taken 26687 times.
✓ Branch 3 taken 2909 times.
|
34641 | } while (fflag && its < 10); |
| 488 | } | ||
| 489 | |||
| 490 | 1230 | static void search_for_pns(AACEncContext *s, AVCodecContext *avctx, SingleChannelElement *sce) | |
| 491 | { | ||
| 492 | FFPsyBand *band; | ||
| 493 | int w, g, w2, i; | ||
| 494 | 1230 | int wlen = 1024 / sce->ics.num_windows; | |
| 495 | int bandwidth, cutoff; | ||
| 496 | 1230 | float *PNS = &s->scoefs[0*128], *PNS34 = &s->scoefs[1*128]; | |
| 497 | 1230 | float *NOR34 = &s->scoefs[3*128]; | |
| 498 | uint8_t nextband[128]; | ||
| 499 | 1230 | const float lambda = s->lambda; | |
| 500 | 1230 | const float freq_mult = avctx->sample_rate*0.5f/wlen; | |
| 501 | 1230 | const float thr_mult = NOISE_LAMBDA_REPLACE*(100.0f/lambda); | |
| 502 |
2/6✗ Branch 0 not taken.
✓ Branch 1 taken 1230 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 1230 times.
✗ Branch 4 not taken.
✗ Branch 5 not taken.
|
1230 | const float spread_threshold = FFMIN(0.75f, NOISE_SPREAD_THRESHOLD*FFMAX(0.5f, lambda/100.f)); |
| 503 | 1230 | const float dist_bias = av_clipf(4.f * 120 / lambda, 0.25f, 4.0f); | |
| 504 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1230 times.
|
1230 | const float pns_transient_energy_r = FFMIN(0.7f, lambda / 140.f); |
| 505 | |||
| 506 | 1230 | int prev = -1000, prev_sf = -1; | |
| 507 | |||
| 508 | /* PNS candidacy must use the coder's actual coding bandwidth (s->bandwidth, | ||
| 509 | * fixed at init), not a separate heuristic, or it evaluates a different band | ||
| 510 | * range than the coder later codes. */ | ||
| 511 | 1230 | bandwidth = s->bandwidth; | |
| 512 | 1230 | cutoff = bandwidth * 2 * wlen / avctx->sample_rate; | |
| 513 | |||
| 514 | 1230 | memcpy(sce->band_alt, sce->band_type, sizeof(sce->band_type)); | |
| 515 | 1230 | ff_init_nextband_map(sce, nextband); | |
| 516 |
2/2✓ Branch 0 taken 1304 times.
✓ Branch 1 taken 1230 times.
|
2534 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 517 | 1304 | int wstart = w*128; | |
| 518 |
2/2✓ Branch 0 taken 56428 times.
✓ Branch 1 taken 1304 times.
|
57732 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 519 | int noise_sfi; | ||
| 520 | 56428 | float dist1 = 0.0f, dist2 = 0.0f, noise_amp; | |
| 521 | 56428 | float pns_energy = 0.0f, pns_tgt_energy, energy_ratio, dist_thresh; | |
| 522 | 56428 | float sfb_energy = 0.0f, threshold = 0.0f, spread = 2.0f; | |
| 523 | 56428 | float min_energy = -1.0f, max_energy = 0.0f; | |
| 524 | 56428 | const int start = wstart+sce->ics.swb_offset[g]; | |
| 525 | 56428 | const float freq = (start-wstart)*freq_mult; | |
| 526 |
2/2✓ Branch 0 taken 28018 times.
✓ Branch 1 taken 28410 times.
|
56428 | const float freq_boost = FFMAX(0.88f*freq/NOISE_LOW_LIMIT, 1.0f); |
| 527 |
4/4✓ Branch 0 taken 29130 times.
✓ Branch 1 taken 27298 times.
✓ Branch 2 taken 2716 times.
✓ Branch 3 taken 26414 times.
|
56428 | if (freq < NOISE_LOW_LIMIT || (start-wstart) >= cutoff) { |
| 528 |
2/2✓ Branch 0 taken 26973 times.
✓ Branch 1 taken 3041 times.
|
30014 | if (!sce->zeroes[w*16+g]) |
| 529 | 26973 | prev_sf = sce->sf_idx[w*16+g]; | |
| 530 | 30014 | continue; | |
| 531 | } | ||
| 532 |
2/2✓ Branch 0 taken 28602 times.
✓ Branch 1 taken 26414 times.
|
55016 | for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) { |
| 533 | 28602 | band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g]; | |
| 534 | 28602 | sfb_energy += band->energy; | |
| 535 |
2/2✓ Branch 0 taken 27004 times.
✓ Branch 1 taken 1598 times.
|
28602 | spread = FFMIN(spread, band->spread); |
| 536 | 28602 | threshold += band->threshold; | |
| 537 |
2/2✓ Branch 0 taken 26414 times.
✓ Branch 1 taken 2188 times.
|
28602 | if (!w2) { |
| 538 | 26414 | min_energy = max_energy = band->energy; | |
| 539 | } else { | ||
| 540 |
2/2✓ Branch 0 taken 779 times.
✓ Branch 1 taken 1409 times.
|
2188 | min_energy = FFMIN(min_energy, band->energy); |
| 541 |
2/2✓ Branch 0 taken 1278 times.
✓ Branch 1 taken 910 times.
|
2188 | max_energy = FFMAX(max_energy, band->energy); |
| 542 | } | ||
| 543 | } | ||
| 544 | |||
| 545 | /* Ramps down at ~8000Hz and loosens the dist threshold */ | ||
| 546 | 26414 | dist_thresh = av_clipf(2.5f*NOISE_LOW_LIMIT/freq, 0.5f, 2.5f) * dist_bias; | |
| 547 | |||
| 548 | /* PNS is acceptable when all of these are true: | ||
| 549 | * 1. high spread energy (noise-like band) | ||
| 550 | * 2. near-threshold energy (high PE means the random nature of PNS content will be noticed) | ||
| 551 | * 3. on short window groups, all windows have similar energy (variations in energy would be destroyed by PNS) | ||
| 552 | * | ||
| 553 | * At this stage, point 2 is relaxed for zeroed bands near the noise threshold (hole avoidance is more important) | ||
| 554 | */ | ||
| 555 |
3/4✓ Branch 0 taken 26186 times.
✓ Branch 1 taken 228 times.
✓ Branch 3 taken 26186 times.
✗ Branch 4 not taken.
|
26414 | if ((!sce->zeroes[w*16+g] && !ff_sfdelta_can_remove_band(sce, nextband, prev_sf, w*16+g)) || |
| 556 |
8/8✓ Branch 0 taken 26186 times.
✓ Branch 1 taken 228 times.
✓ Branch 2 taken 3583 times.
✓ Branch 3 taken 22603 times.
✓ Branch 4 taken 3725 times.
✓ Branch 5 taken 86 times.
✓ Branch 6 taken 25978 times.
✓ Branch 7 taken 350 times.
|
26414 | ((sce->zeroes[w*16+g] || !sce->band_alt[w*16+g]) && sfb_energy < threshold*sqrtf(1.0f/freq_boost)) || spread < spread_threshold || |
| 557 |
5/6✓ Branch 0 taken 25978 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 22397 times.
✓ Branch 3 taken 3581 times.
✓ Branch 4 taken 12888 times.
✓ Branch 5 taken 9509 times.
|
25978 | (!sce->zeroes[w*16+g] && sce->band_alt[w*16+g] && sfb_energy > threshold*thr_mult*freq_boost) || |
| 558 |
2/2✓ Branch 0 taken 207 times.
✓ Branch 1 taken 16262 times.
|
16469 | min_energy < pns_transient_energy_r * max_energy ) { |
| 559 | 10152 | sce->pns_ener[w*16+g] = sfb_energy; | |
| 560 |
2/2✓ Branch 0 taken 9924 times.
✓ Branch 1 taken 228 times.
|
10152 | if (!sce->zeroes[w*16+g]) |
| 561 | 9924 | prev_sf = sce->sf_idx[w*16+g]; | |
| 562 | 10152 | continue; | |
| 563 | } | ||
| 564 | |||
| 565 |
2/2✓ Branch 0 taken 15627 times.
✓ Branch 1 taken 635 times.
|
16262 | pns_tgt_energy = sfb_energy*FFMIN(1.0f, spread*spread); |
| 566 | 16262 | noise_sfi = av_clip(roundf(log2f(pns_tgt_energy)*2), -100, 155); /* Quantize */ | |
| 567 | 16262 | noise_amp = -ff_aac_pow2sf_tab[noise_sfi + POW_SF2_ZERO]; /* Dequantize */ | |
| 568 |
2/2✓ Branch 0 taken 10793 times.
✓ Branch 1 taken 5469 times.
|
16262 | if (prev != -1000) { |
| 569 | 10793 | int noise_sfdiff = noise_sfi - prev + SCALE_DIFF_ZERO; | |
| 570 |
2/4✓ Branch 0 taken 10793 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 10793 times.
|
10793 | if (noise_sfdiff < 0 || noise_sfdiff > 2*SCALE_MAX_DIFF) { |
| 571 | ✗ | if (!sce->zeroes[w*16+g]) | |
| 572 | ✗ | prev_sf = sce->sf_idx[w*16+g]; | |
| 573 | ✗ | continue; | |
| 574 | } | ||
| 575 | } | ||
| 576 |
2/2✓ Branch 0 taken 16322 times.
✓ Branch 1 taken 16262 times.
|
32584 | for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) { |
| 577 | float band_energy, scale, pns_senergy; | ||
| 578 | 16322 | const int start_c = (w+w2)*128+sce->ics.swb_offset[g]; | |
| 579 | 16322 | band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g]; | |
| 580 |
2/2✓ Branch 0 taken 536788 times.
✓ Branch 1 taken 16322 times.
|
553110 | for (i = 0; i < sce->ics.swb_sizes[g]; i++) { |
| 581 | 536788 | s->random_state = lcg_random(s->random_state); | |
| 582 | 536788 | PNS[i] = s->random_state; | |
| 583 | } | ||
| 584 | 16322 | band_energy = s->fdsp->scalarproduct_float(PNS, PNS, sce->ics.swb_sizes[g]); | |
| 585 | 16322 | scale = noise_amp/sqrtf(band_energy); | |
| 586 | 16322 | s->fdsp->vector_fmul_scalar(PNS, PNS, scale, sce->ics.swb_sizes[g]); | |
| 587 | 16322 | pns_senergy = s->fdsp->scalarproduct_float(PNS, PNS, sce->ics.swb_sizes[g]); | |
| 588 | 16322 | pns_energy += pns_senergy; | |
| 589 | 16322 | s->aacdsp.abs_pow34(NOR34, &sce->coeffs[start_c], sce->ics.swb_sizes[g]); | |
| 590 | 16322 | s->aacdsp.abs_pow34(PNS34, PNS, sce->ics.swb_sizes[g]); | |
| 591 | 32644 | dist1 += quantize_band_cost(s, &sce->coeffs[start_c], | |
| 592 | NOR34, | ||
| 593 | 16322 | sce->ics.swb_sizes[g], | |
| 594 | 16322 | sce->sf_idx[(w+w2)*16+g], | |
| 595 | 16322 | sce->band_alt[(w+w2)*16+g], | |
| 596 | 16322 | lambda/band->threshold, INFINITY, NULL, NULL); | |
| 597 | /* Estimate rd on average as 5 bits for SF, 4 for the CB, plus spread energy * lambda/thr */ | ||
| 598 | 16322 | dist2 += band->energy/(band->spread*band->spread)*lambda*dist_thresh/band->threshold; | |
| 599 | } | ||
| 600 |
3/4✓ Branch 0 taken 16262 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 2682 times.
✓ Branch 3 taken 13580 times.
|
16262 | if (g && sce->band_type[w*16+g-1] == NOISE_BT) { |
| 601 | 2682 | dist2 += 5; | |
| 602 | } else { | ||
| 603 | 13580 | dist2 += 9; | |
| 604 | } | ||
| 605 | 16262 | energy_ratio = pns_tgt_energy/pns_energy; /* Compensates for quantization error */ | |
| 606 | 16262 | sce->pns_ener[w*16+g] = energy_ratio*pns_tgt_energy; | |
| 607 |
7/10✓ Branch 0 taken 16262 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 12874 times.
✓ Branch 3 taken 3388 times.
✓ Branch 4 taken 12502 times.
✓ Branch 5 taken 372 times.
✓ Branch 6 taken 12502 times.
✗ Branch 7 not taken.
✗ Branch 8 not taken.
✓ Branch 9 taken 12502 times.
|
16262 | if (sce->zeroes[w*16+g] || !sce->band_alt[w*16+g] || (energy_ratio > 0.85f && energy_ratio < 1.25f && dist2 < dist1)) { |
| 608 | 3388 | sce->band_type[w*16+g] = NOISE_BT; | |
| 609 | 3388 | sce->zeroes[w*16+g] = 0; | |
| 610 | 3388 | prev = noise_sfi; | |
| 611 | } else { | ||
| 612 |
1/2✓ Branch 0 taken 12874 times.
✗ Branch 1 not taken.
|
12874 | if (!sce->zeroes[w*16+g]) |
| 613 | 12874 | prev_sf = sce->sf_idx[w*16+g]; | |
| 614 | } | ||
| 615 | } | ||
| 616 | } | ||
| 617 | 1230 | } | |
| 618 | |||
| 619 | 4638 | static void mark_pns(AACEncContext *s, AVCodecContext *avctx, SingleChannelElement *sce) | |
| 620 | { | ||
| 621 | FFPsyBand *band; | ||
| 622 | int w, g, w2; | ||
| 623 | 4638 | int wlen = 1024 / sce->ics.num_windows; | |
| 624 | int bandwidth, cutoff; | ||
| 625 | 4638 | const float lambda = s->lambda; | |
| 626 | 4638 | const float freq_mult = avctx->sample_rate*0.5f/wlen; | |
| 627 |
2/6✗ Branch 0 not taken.
✓ Branch 1 taken 4638 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 4638 times.
✗ Branch 4 not taken.
✗ Branch 5 not taken.
|
4638 | const float spread_threshold = FFMIN(0.75f, NOISE_SPREAD_THRESHOLD*FFMAX(0.5f, lambda/100.f)); |
| 628 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 4638 times.
|
4638 | const float pns_transient_energy_r = FFMIN(0.7f, lambda / 140.f); |
| 629 | |||
| 630 | /* PNS candidacy must use the coder's actual coding bandwidth (s->bandwidth, | ||
| 631 | * fixed at init), not a separate heuristic, or it evaluates a different band | ||
| 632 | * range than the coder later codes (NMR relies on this output directly). */ | ||
| 633 | 4638 | bandwidth = s->bandwidth; | |
| 634 | 4638 | cutoff = bandwidth * 2 * wlen / avctx->sample_rate; | |
| 635 | |||
| 636 | 4638 | memcpy(sce->band_alt, sce->band_type, sizeof(sce->band_type)); | |
| 637 |
2/2✓ Branch 0 taken 4975 times.
✓ Branch 1 taken 4638 times.
|
9613 | for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { |
| 638 |
2/2✓ Branch 0 taken 215437 times.
✓ Branch 1 taken 4975 times.
|
220412 | for (g = 0; g < sce->ics.num_swb; g++) { |
| 639 | 215437 | float sfb_energy = 0.0f, threshold = 0.0f, spread = 2.0f; | |
| 640 | 215437 | float min_energy = -1.0f, max_energy = 0.0f; | |
| 641 | 215437 | const int start = sce->ics.swb_offset[g]; | |
| 642 | 215437 | const float freq = start*freq_mult; | |
| 643 |
2/2✓ Branch 0 taken 106314 times.
✓ Branch 1 taken 109123 times.
|
215437 | const float freq_boost = FFMAX(0.88f*freq/NOISE_LOW_LIMIT, 1.0f); |
| 644 |
4/4✓ Branch 0 taken 110529 times.
✓ Branch 1 taken 104908 times.
✓ Branch 2 taken 19738 times.
✓ Branch 3 taken 90791 times.
|
215437 | if (freq < NOISE_LOW_LIMIT || start >= cutoff) { |
| 645 | 124646 | sce->can_pns[w*16+g] = 0; | |
| 646 | 124646 | continue; | |
| 647 | } | ||
| 648 |
2/2✓ Branch 0 taken 101055 times.
✓ Branch 1 taken 90791 times.
|
191846 | for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) { |
| 649 | 101055 | band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g]; | |
| 650 | 101055 | sfb_energy += band->energy; | |
| 651 |
2/2✓ Branch 0 taken 93715 times.
✓ Branch 1 taken 7340 times.
|
101055 | spread = FFMIN(spread, band->spread); |
| 652 | 101055 | threshold += band->threshold; | |
| 653 |
2/2✓ Branch 0 taken 90791 times.
✓ Branch 1 taken 10264 times.
|
101055 | if (!w2) { |
| 654 | 90791 | min_energy = max_energy = band->energy; | |
| 655 | } else { | ||
| 656 |
2/2✓ Branch 0 taken 3630 times.
✓ Branch 1 taken 6634 times.
|
10264 | min_energy = FFMIN(min_energy, band->energy); |
| 657 |
2/2✓ Branch 0 taken 5743 times.
✓ Branch 1 taken 4521 times.
|
10264 | max_energy = FFMAX(max_energy, band->energy); |
| 658 | } | ||
| 659 | } | ||
| 660 | |||
| 661 | /* PNS is acceptable when all of these are true: | ||
| 662 | * 1. high spread energy (noise-like band) | ||
| 663 | * 2. near-threshold energy (high PE means the random nature of PNS content will be noticed) | ||
| 664 | * 3. on short window groups, all windows have similar energy (variations in energy would be destroyed by PNS) | ||
| 665 | */ | ||
| 666 | 90791 | sce->pns_ener[w*16+g] = sfb_energy; | |
| 667 | { | ||
| 668 | /* near-mask PNS class (E in [thr/4, 2*thr]): deletion | ||
| 669 | * candidates go to noise, not silence (AAC_PNSHOLE) */ | ||
| 670 |
2/2✓ Branch 0 taken 10050 times.
✓ Branch 1 taken 80741 times.
|
100841 | int near = sfb_energy < 2.0f * threshold && |
| 671 |
2/2✓ Branch 0 taken 9880 times.
✓ Branch 1 taken 170 times.
|
10050 | sfb_energy > threshold * 0.25f; |
| 672 |
2/2✓ Branch 0 taken 9880 times.
✓ Branch 1 taken 80911 times.
|
90791 | if (near) { |
| 673 | /* deletion candidate: noise beats the ~silent rendition */ | ||
| 674 |
2/2✓ Branch 0 taken 9748 times.
✓ Branch 1 taken 132 times.
|
19628 | sce->can_pns[w*16+g] = spread >= spread_threshold && |
| 675 |
1/2✓ Branch 0 taken 9748 times.
✗ Branch 1 not taken.
|
9748 | min_energy >= 0.2f * max_energy; |
| 676 |
6/6✓ Branch 0 taken 80741 times.
✓ Branch 1 taken 170 times.
✓ Branch 2 taken 80034 times.
✓ Branch 3 taken 707 times.
✓ Branch 4 taken 3489 times.
✓ Branch 5 taken 76545 times.
|
80911 | } else if (sfb_energy < threshold*sqrtf(1.5f/freq_boost) || spread < spread_threshold || min_energy < pns_transient_energy_r * max_energy) { |
| 677 | 4366 | sce->can_pns[w*16+g] = 0; | |
| 678 | } else { | ||
| 679 | 76545 | sce->can_pns[w*16+g] = 1; | |
| 680 | } | ||
| 681 | } | ||
| 682 | } | ||
| 683 | } | ||
| 684 | 4638 | } | |
| 685 | |||
| 686 | ✗ | static void search_for_ms(AACEncContext *s, ChannelElement *cpe) | |
| 687 | { | ||
| 688 | ✗ | int start = 0, i, w, w2, g, sid_sf_boost, prev_mid, prev_side; | |
| 689 | uint8_t nextband0[128], nextband1[128]; | ||
| 690 | ✗ | float *M = s->scoefs + 128*0, *S = s->scoefs + 128*1; | |
| 691 | ✗ | float *L34 = s->scoefs + 128*2, *R34 = s->scoefs + 128*3; | |
| 692 | ✗ | float *M34 = s->scoefs + 128*4, *S34 = s->scoefs + 128*5; | |
| 693 | ✗ | const float lambda = s->lambda; | |
| 694 | ✗ | const float mslambda = FFMIN(1.0f, lambda / 120.f); | |
| 695 | ✗ | SingleChannelElement *sce0 = &cpe->ch[0]; | |
| 696 | ✗ | SingleChannelElement *sce1 = &cpe->ch[1]; | |
| 697 | ✗ | if (!cpe->common_window) | |
| 698 | ✗ | return; | |
| 699 | |||
| 700 | /** Scout out next nonzero bands */ | ||
| 701 | ✗ | ff_init_nextband_map(sce0, nextband0); | |
| 702 | ✗ | ff_init_nextband_map(sce1, nextband1); | |
| 703 | |||
| 704 | ✗ | prev_mid = sce0->sf_idx[0]; | |
| 705 | ✗ | prev_side = sce1->sf_idx[0]; | |
| 706 | ✗ | for (w = 0; w < sce0->ics.num_windows; w += sce0->ics.group_len[w]) { | |
| 707 | ✗ | start = 0; | |
| 708 | ✗ | for (g = 0; g < sce0->ics.num_swb; g++) { | |
| 709 | ✗ | float bmax = bval2bmax(g * 17.0f / sce0->ics.num_swb) / 0.0045f; | |
| 710 | ✗ | if (!cpe->is_mask[w*16+g]) | |
| 711 | ✗ | cpe->ms_mask[w*16+g] = 0; | |
| 712 | ✗ | if (!sce0->zeroes[w*16+g] && !sce1->zeroes[w*16+g] && !cpe->is_mask[w*16+g]) { | |
| 713 | ✗ | float Mmax = 0.0f, Smax = 0.0f; | |
| 714 | |||
| 715 | /* Must compute mid/side SF and book for the whole window group */ | ||
| 716 | ✗ | for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) { | |
| 717 | ✗ | for (i = 0; i < sce0->ics.swb_sizes[g]; i++) { | |
| 718 | ✗ | M[i] = (sce0->coeffs[start+(w+w2)*128+i] | |
| 719 | ✗ | + sce1->coeffs[start+(w+w2)*128+i]) * 0.5; | |
| 720 | ✗ | S[i] = M[i] | |
| 721 | ✗ | - sce1->coeffs[start+(w+w2)*128+i]; | |
| 722 | } | ||
| 723 | ✗ | s->aacdsp.abs_pow34(M34, M, sce0->ics.swb_sizes[g]); | |
| 724 | ✗ | s->aacdsp.abs_pow34(S34, S, sce0->ics.swb_sizes[g]); | |
| 725 | ✗ | for (i = 0; i < sce0->ics.swb_sizes[g]; i++ ) { | |
| 726 | ✗ | Mmax = FFMAX(Mmax, M34[i]); | |
| 727 | ✗ | Smax = FFMAX(Smax, S34[i]); | |
| 728 | } | ||
| 729 | } | ||
| 730 | |||
| 731 | ✗ | for (sid_sf_boost = 0; sid_sf_boost < 4; sid_sf_boost++) { | |
| 732 | ✗ | float dist1 = 0.0f, dist2 = 0.0f; | |
| 733 | ✗ | int B0 = 0, B1 = 0; | |
| 734 | int minidx; | ||
| 735 | int mididx, sididx; | ||
| 736 | int midcb, sidcb; | ||
| 737 | |||
| 738 | ✗ | minidx = FFMIN(sce0->sf_idx[w*16+g], sce1->sf_idx[w*16+g]); | |
| 739 | ✗ | mididx = av_clip(minidx, 0, SCALE_MAX_POS - SCALE_DIV_512); | |
| 740 | ✗ | sididx = av_clip(minidx - sid_sf_boost * 3, 0, SCALE_MAX_POS - SCALE_DIV_512); | |
| 741 | ✗ | if (sce0->band_type[w*16+g] != NOISE_BT && sce1->band_type[w*16+g] != NOISE_BT | |
| 742 | ✗ | && ( !ff_sfdelta_can_replace(sce0, nextband0, prev_mid, mididx, w*16+g) | |
| 743 | ✗ | || !ff_sfdelta_can_replace(sce1, nextband1, prev_side, sididx, w*16+g))) { | |
| 744 | /* scalefactor range violation, bad stuff, will decrease quality unacceptably */ | ||
| 745 | ✗ | continue; | |
| 746 | } | ||
| 747 | |||
| 748 | ✗ | midcb = find_min_book(Mmax, mididx); | |
| 749 | ✗ | sidcb = find_min_book(Smax, sididx); | |
| 750 | |||
| 751 | /* No CB can be zero */ | ||
| 752 | ✗ | midcb = FFMAX(1,midcb); | |
| 753 | ✗ | sidcb = FFMAX(1,sidcb); | |
| 754 | |||
| 755 | ✗ | for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) { | |
| 756 | ✗ | FFPsyBand *band0 = &s->psy.ch[s->cur_channel+0].psy_bands[(w+w2)*16+g]; | |
| 757 | ✗ | FFPsyBand *band1 = &s->psy.ch[s->cur_channel+1].psy_bands[(w+w2)*16+g]; | |
| 758 | ✗ | float minthr = FFMIN(band0->threshold, band1->threshold); | |
| 759 | int b1,b2,b3,b4; | ||
| 760 | ✗ | for (i = 0; i < sce0->ics.swb_sizes[g]; i++) { | |
| 761 | ✗ | M[i] = (sce0->coeffs[start+(w+w2)*128+i] | |
| 762 | ✗ | + sce1->coeffs[start+(w+w2)*128+i]) * 0.5; | |
| 763 | ✗ | S[i] = M[i] | |
| 764 | ✗ | - sce1->coeffs[start+(w+w2)*128+i]; | |
| 765 | } | ||
| 766 | |||
| 767 | ✗ | s->aacdsp.abs_pow34(L34, sce0->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]); | |
| 768 | ✗ | s->aacdsp.abs_pow34(R34, sce1->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]); | |
| 769 | ✗ | s->aacdsp.abs_pow34(M34, M, sce0->ics.swb_sizes[g]); | |
| 770 | ✗ | s->aacdsp.abs_pow34(S34, S, sce0->ics.swb_sizes[g]); | |
| 771 | ✗ | dist1 += quantize_band_cost(s, &sce0->coeffs[start + (w+w2)*128], | |
| 772 | L34, | ||
| 773 | ✗ | sce0->ics.swb_sizes[g], | |
| 774 | ✗ | sce0->sf_idx[w*16+g], | |
| 775 | ✗ | sce0->band_type[w*16+g], | |
| 776 | ✗ | lambda / (band0->threshold + FLT_MIN), INFINITY, &b1, NULL); | |
| 777 | ✗ | dist1 += quantize_band_cost(s, &sce1->coeffs[start + (w+w2)*128], | |
| 778 | R34, | ||
| 779 | ✗ | sce1->ics.swb_sizes[g], | |
| 780 | ✗ | sce1->sf_idx[w*16+g], | |
| 781 | ✗ | sce1->band_type[w*16+g], | |
| 782 | ✗ | lambda / (band1->threshold + FLT_MIN), INFINITY, &b2, NULL); | |
| 783 | ✗ | dist2 += quantize_band_cost(s, M, | |
| 784 | M34, | ||
| 785 | ✗ | sce0->ics.swb_sizes[g], | |
| 786 | mididx, | ||
| 787 | midcb, | ||
| 788 | ✗ | lambda / (minthr + FLT_MIN), INFINITY, &b3, NULL); | |
| 789 | ✗ | dist2 += quantize_band_cost(s, S, | |
| 790 | S34, | ||
| 791 | ✗ | sce1->ics.swb_sizes[g], | |
| 792 | sididx, | ||
| 793 | sidcb, | ||
| 794 | ✗ | mslambda / (minthr * bmax + FLT_MIN), INFINITY, &b4, NULL); | |
| 795 | ✗ | B0 += b1+b2; | |
| 796 | ✗ | B1 += b3+b4; | |
| 797 | ✗ | dist1 -= b1+b2; | |
| 798 | ✗ | dist2 -= b3+b4; | |
| 799 | } | ||
| 800 | ✗ | cpe->ms_mask[w*16+g] = dist2 <= dist1 && B1 < B0; | |
| 801 | ✗ | if (cpe->ms_mask[w*16+g]) { | |
| 802 | ✗ | if (sce0->band_type[w*16+g] != NOISE_BT && sce1->band_type[w*16+g] != NOISE_BT) { | |
| 803 | ✗ | sce0->sf_idx[w*16+g] = mididx; | |
| 804 | ✗ | sce1->sf_idx[w*16+g] = sididx; | |
| 805 | ✗ | sce0->band_type[w*16+g] = midcb; | |
| 806 | ✗ | sce1->band_type[w*16+g] = sidcb; | |
| 807 | ✗ | } else if ((sce0->band_type[w*16+g] != NOISE_BT) ^ (sce1->band_type[w*16+g] != NOISE_BT)) { | |
| 808 | /* ms_mask unneeded, and it confuses some decoders */ | ||
| 809 | ✗ | cpe->ms_mask[w*16+g] = 0; | |
| 810 | } | ||
| 811 | ✗ | break; | |
| 812 | ✗ | } else if (B1 > B0) { | |
| 813 | /* More boost won't fix this */ | ||
| 814 | ✗ | break; | |
| 815 | } | ||
| 816 | } | ||
| 817 | } | ||
| 818 | ✗ | if (!sce0->zeroes[w*16+g] && sce0->band_type[w*16+g] < RESERVED_BT) | |
| 819 | ✗ | prev_mid = sce0->sf_idx[w*16+g]; | |
| 820 | ✗ | if (!sce1->zeroes[w*16+g] && !cpe->is_mask[w*16+g] && sce1->band_type[w*16+g] < RESERVED_BT) | |
| 821 | ✗ | prev_side = sce1->sf_idx[w*16+g]; | |
| 822 | ✗ | start += sce0->ics.swb_sizes[g]; | |
| 823 | } | ||
| 824 | } | ||
| 825 | } | ||
| 826 | |||
| 827 | const AACCoefficientsEncoder ff_aac_coders[AAC_CODER_NB] = { | ||
| 828 | [AAC_CODER_TWOLOOP] = { | ||
| 829 | search_for_quantizers_twoloop, | ||
| 830 | codebook_trellis_rate, | ||
| 831 | quantize_and_encode_band, | ||
| 832 | ff_aac_encode_tns_info, | ||
| 833 | ff_aac_apply_tns, | ||
| 834 | set_special_band_scalefactors, | ||
| 835 | search_for_pns, | ||
| 836 | mark_pns, | ||
| 837 | ff_aac_search_for_tns, | ||
| 838 | search_for_ms, | ||
| 839 | ff_aac_search_for_is, | ||
| 840 | }, | ||
| 841 | [AAC_CODER_FAST] = { | ||
| 842 | search_for_quantizers_fast, | ||
| 843 | codebook_trellis_rate, | ||
| 844 | quantize_and_encode_band, | ||
| 845 | ff_aac_encode_tns_info, | ||
| 846 | ff_aac_apply_tns, | ||
| 847 | set_special_band_scalefactors, | ||
| 848 | search_for_pns, | ||
| 849 | mark_pns, | ||
| 850 | ff_aac_search_for_tns, | ||
| 851 | search_for_ms, | ||
| 852 | ff_aac_search_for_is, | ||
| 853 | }, | ||
| 854 | [AAC_CODER_NMR] = { | ||
| 855 | search_for_quantizers_nmr, | ||
| 856 | codebook_trellis_rate, | ||
| 857 | quantize_and_encode_band, | ||
| 858 | ff_aac_encode_tns_info, | ||
| 859 | ff_aac_apply_tns, | ||
| 860 | set_special_band_scalefactors, | ||
| 861 | NULL, /* PNS decided in the trellis (search_for_quantizers_nmr) */ | ||
| 862 | mark_pns, | ||
| 863 | ff_aac_search_for_tns, | ||
| 864 | NULL, | ||
| 865 | NULL, | ||
| 866 | }, | ||
| 867 | }; | ||
| 868 |