FFmpeg  4.0
aacenc_quantization.h
Go to the documentation of this file.
1 /*
2  * AAC encoder quantizer
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 quantizer
25  * @author Rostislav Pehlivanov ( atomnuker gmail com )
26  */
27 
28 #ifndef AVCODEC_AACENC_QUANTIZATION_H
29 #define AVCODEC_AACENC_QUANTIZATION_H
30 
31 #include "aactab.h"
32 #include "aacenc.h"
33 #include "aacenctab.h"
34 #include "aacenc_utils.h"
35 
36 /**
37  * Calculate rate distortion cost for quantizing with given codebook
38  *
39  * @return quantization distortion
40  */
42  struct AACEncContext *s,
43  PutBitContext *pb, const float *in, float *out,
44  const float *scaled, int size, int scale_idx,
45  int cb, const float lambda, const float uplim,
46  int *bits, float *energy, int BT_ZERO, int BT_UNSIGNED,
47  int BT_PAIR, int BT_ESC, int BT_NOISE, int BT_STEREO,
48  const float ROUNDING)
49 {
50  const int q_idx = POW_SF2_ZERO - scale_idx + SCALE_ONE_POS - SCALE_DIV_512;
51  const float Q = ff_aac_pow2sf_tab [q_idx];
52  const float Q34 = ff_aac_pow34sf_tab[q_idx];
53  const float IQ = ff_aac_pow2sf_tab [POW_SF2_ZERO + scale_idx - SCALE_ONE_POS + SCALE_DIV_512];
54  const float CLIPPED_ESCAPE = 165140.0f*IQ;
55  int i, j;
56  float cost = 0;
57  float qenergy = 0;
58  const int dim = BT_PAIR ? 2 : 4;
59  int resbits = 0;
60  int off;
61 
62  if (BT_ZERO || BT_NOISE || BT_STEREO) {
63  for (i = 0; i < size; i++)
64  cost += in[i]*in[i];
65  if (bits)
66  *bits = 0;
67  if (energy)
68  *energy = qenergy;
69  if (out) {
70  for (i = 0; i < size; i += dim)
71  for (j = 0; j < dim; j++)
72  out[i+j] = 0.0f;
73  }
74  return cost * lambda;
75  }
76  if (!scaled) {
77  s->abs_pow34(s->scoefs, in, size);
78  scaled = s->scoefs;
79  }
80  s->quant_bands(s->qcoefs, in, scaled, size, !BT_UNSIGNED, aac_cb_maxval[cb], Q34, ROUNDING);
81  if (BT_UNSIGNED) {
82  off = 0;
83  } else {
84  off = aac_cb_maxval[cb];
85  }
86  for (i = 0; i < size; i += dim) {
87  const float *vec;
88  int *quants = s->qcoefs + i;
89  int curidx = 0;
90  int curbits;
91  float quantized, rd = 0.0f;
92  for (j = 0; j < dim; j++) {
93  curidx *= aac_cb_range[cb];
94  curidx += quants[j] + off;
95  }
96  curbits = ff_aac_spectral_bits[cb-1][curidx];
97  vec = &ff_aac_codebook_vectors[cb-1][curidx*dim];
98  if (BT_UNSIGNED) {
99  for (j = 0; j < dim; j++) {
100  float t = fabsf(in[i+j]);
101  float di;
102  if (BT_ESC && vec[j] == 64.0f) { //FIXME: slow
103  if (t >= CLIPPED_ESCAPE) {
104  quantized = CLIPPED_ESCAPE;
105  curbits += 21;
106  } else {
107  int c = av_clip_uintp2(quant(t, Q, ROUNDING), 13);
108  quantized = c*cbrtf(c)*IQ;
109  curbits += av_log2(c)*2 - 4 + 1;
110  }
111  } else {
112  quantized = vec[j]*IQ;
113  }
114  di = t - quantized;
115  if (out)
116  out[i+j] = in[i+j] >= 0 ? quantized : -quantized;
117  if (vec[j] != 0.0f)
118  curbits++;
119  qenergy += quantized*quantized;
120  rd += di*di;
121  }
122  } else {
123  for (j = 0; j < dim; j++) {
124  quantized = vec[j]*IQ;
125  qenergy += quantized*quantized;
126  if (out)
127  out[i+j] = quantized;
128  rd += (in[i+j] - quantized)*(in[i+j] - quantized);
129  }
130  }
131  cost += rd * lambda + curbits;
132  resbits += curbits;
133  if (cost >= uplim)
134  return uplim;
135  if (pb) {
136  put_bits(pb, ff_aac_spectral_bits[cb-1][curidx], ff_aac_spectral_codes[cb-1][curidx]);
137  if (BT_UNSIGNED)
138  for (j = 0; j < dim; j++)
139  if (ff_aac_codebook_vectors[cb-1][curidx*dim+j] != 0.0f)
140  put_bits(pb, 1, in[i+j] < 0.0f);
141  if (BT_ESC) {
142  for (j = 0; j < 2; j++) {
143  if (ff_aac_codebook_vectors[cb-1][curidx*2+j] == 64.0f) {
144  int coef = av_clip_uintp2(quant(fabsf(in[i+j]), Q, ROUNDING), 13);
145  int len = av_log2(coef);
146 
147  put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2);
148  put_sbits(pb, len, coef);
149  }
150  }
151  }
152  }
153  }
154 
155  if (bits)
156  *bits = resbits;
157  if (energy)
158  *energy = qenergy;
159  return cost;
160 }
161 
163  const float *in, float *quant, const float *scaled,
164  int size, int scale_idx, int cb,
165  const float lambda, const float uplim,
166  int *bits, float *energy) {
167  av_assert0(0);
168  return 0.0f;
169 }
170 
171 #define QUANTIZE_AND_ENCODE_BAND_COST_FUNC(NAME, BT_ZERO, BT_UNSIGNED, BT_PAIR, BT_ESC, BT_NOISE, BT_STEREO, ROUNDING) \
172 static float quantize_and_encode_band_cost_ ## NAME( \
173  struct AACEncContext *s, \
174  PutBitContext *pb, const float *in, float *quant, \
175  const float *scaled, int size, int scale_idx, \
176  int cb, const float lambda, const float uplim, \
177  int *bits, float *energy) { \
178  return quantize_and_encode_band_cost_template( \
179  s, pb, in, quant, scaled, size, scale_idx, \
180  BT_ESC ? ESC_BT : cb, lambda, uplim, bits, energy, \
181  BT_ZERO, BT_UNSIGNED, BT_PAIR, BT_ESC, BT_NOISE, BT_STEREO, \
182  ROUNDING); \
183 }
184 
185 QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ZERO, 1, 0, 0, 0, 0, 0, ROUND_STANDARD)
186 QUANTIZE_AND_ENCODE_BAND_COST_FUNC(SQUAD, 0, 0, 0, 0, 0, 0, ROUND_STANDARD)
187 QUANTIZE_AND_ENCODE_BAND_COST_FUNC(UQUAD, 0, 1, 0, 0, 0, 0, ROUND_STANDARD)
188 QUANTIZE_AND_ENCODE_BAND_COST_FUNC(SPAIR, 0, 0, 1, 0, 0, 0, ROUND_STANDARD)
189 QUANTIZE_AND_ENCODE_BAND_COST_FUNC(UPAIR, 0, 1, 1, 0, 0, 0, ROUND_STANDARD)
191 QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ESC_RTZ, 0, 1, 1, 1, 0, 0, ROUND_TO_ZERO)
192 QUANTIZE_AND_ENCODE_BAND_COST_FUNC(NOISE, 0, 0, 0, 0, 1, 0, ROUND_STANDARD)
194 
195 static float (*const quantize_and_encode_band_cost_arr[])(
196  struct AACEncContext *s,
197  PutBitContext *pb, const float *in, float *quant,
198  const float *scaled, int size, int scale_idx,
199  int cb, const float lambda, const float uplim,
200  int *bits, float *energy) = {
201  quantize_and_encode_band_cost_ZERO,
202  quantize_and_encode_band_cost_SQUAD,
203  quantize_and_encode_band_cost_SQUAD,
204  quantize_and_encode_band_cost_UQUAD,
205  quantize_and_encode_band_cost_UQUAD,
206  quantize_and_encode_band_cost_SPAIR,
207  quantize_and_encode_band_cost_SPAIR,
208  quantize_and_encode_band_cost_UPAIR,
209  quantize_and_encode_band_cost_UPAIR,
210  quantize_and_encode_band_cost_UPAIR,
211  quantize_and_encode_band_cost_UPAIR,
212  quantize_and_encode_band_cost_ESC,
213  quantize_and_encode_band_cost_NONE, /* CB 12 doesn't exist */
214  quantize_and_encode_band_cost_NOISE,
215  quantize_and_encode_band_cost_STEREO,
216  quantize_and_encode_band_cost_STEREO,
217 };
218 
220  struct AACEncContext *s,
221  PutBitContext *pb, const float *in, float *quant,
222  const float *scaled, int size, int scale_idx,
223  int cb, const float lambda, const float uplim,
224  int *bits, float *energy) = {
225  quantize_and_encode_band_cost_ZERO,
226  quantize_and_encode_band_cost_SQUAD,
227  quantize_and_encode_band_cost_SQUAD,
228  quantize_and_encode_band_cost_UQUAD,
229  quantize_and_encode_band_cost_UQUAD,
230  quantize_and_encode_band_cost_SPAIR,
231  quantize_and_encode_band_cost_SPAIR,
232  quantize_and_encode_band_cost_UPAIR,
233  quantize_and_encode_band_cost_UPAIR,
234  quantize_and_encode_band_cost_UPAIR,
235  quantize_and_encode_band_cost_UPAIR,
236  quantize_and_encode_band_cost_ESC_RTZ,
237  quantize_and_encode_band_cost_NONE, /* CB 12 doesn't exist */
238  quantize_and_encode_band_cost_NOISE,
239  quantize_and_encode_band_cost_STEREO,
240  quantize_and_encode_band_cost_STEREO,
241 };
242 
243 #define quantize_and_encode_band_cost( \
244  s, pb, in, quant, scaled, size, scale_idx, cb, \
245  lambda, uplim, bits, energy, rtz) \
246  ((rtz) ? quantize_and_encode_band_cost_rtz_arr : quantize_and_encode_band_cost_arr)[cb]( \
247  s, pb, in, quant, scaled, size, scale_idx, cb, \
248  lambda, uplim, bits, energy)
249 
250 static inline float quantize_band_cost(struct AACEncContext *s, const float *in,
251  const float *scaled, int size, int scale_idx,
252  int cb, const float lambda, const float uplim,
253  int *bits, float *energy, int rtz)
254 {
255  return quantize_and_encode_band_cost(s, NULL, in, NULL, scaled, size, scale_idx,
256  cb, lambda, uplim, bits, energy, rtz);
257 }
258 
259 static inline int quantize_band_cost_bits(struct AACEncContext *s, const float *in,
260  const float *scaled, int size, int scale_idx,
261  int cb, const float lambda, const float uplim,
262  int *bits, float *energy, int rtz)
263 {
264  int auxbits;
265  quantize_and_encode_band_cost(s, NULL, in, NULL, scaled, size, scale_idx,
266  cb, 0.0f, uplim, &auxbits, energy, rtz);
267  if (bits) {
268  *bits = auxbits;
269  }
270  return auxbits;
271 }
272 
274  const float *in, float *out, int size, int scale_idx,
275  int cb, const float lambda, int rtz)
276 {
277  quantize_and_encode_band_cost(s, pb, in, out, NULL, size, scale_idx, cb, lambda,
278  INFINITY, NULL, NULL, rtz);
279 }
280 
282 
283 #endif /* AVCODEC_AACENC_QUANTIZATION_H */
#define NULL
Definition: coverity.c:32
const char * s
Definition: avisynth_c.h:768
int size
static void put_sbits(PutBitContext *pb, int n, int32_t value)
Definition: put_bits.h:240
static void put_bits(Jpeg2000EncoderContext *s, int val, int n)
put n times val bit
Definition: j2kenc.c:207
float lambda
Definition: aacenc.h:400
#define ROUND_TO_ZERO
Definition: aacenc_utils.h:37
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:37
static double cb(void *priv, double x, double y)
Definition: vf_geq.c:112
AAC encoder context.
Definition: aacenc.h:376
int qcoefs[96]
quantized coefficients
Definition: aacenc.h:407
#define quantize_and_encode_band_cost( s, pb, in, quant, scaled, size, scale_idx, cb, lambda, uplim, bits, energy, rtz)
static float(*const quantize_and_encode_band_cost_arr[])(struct AACEncContext *s, PutBitContext *pb, const float *in, float *quant, const float *scaled, int size, int scale_idx, int cb, const float lambda, const float uplim, int *bits, float *energy)
static float(*const quantize_and_encode_band_cost_rtz_arr[])(struct AACEncContext *s, PutBitContext *pb, const float *in, float *quant, const float *scaled, int size, int scale_idx, int cb, const float lambda, const float uplim, int *bits, float *energy)
const float *const ff_aac_codebook_vectors[]
Definition: aactab.c:918
static av_always_inline float quantize_and_encode_band_cost_template(struct AACEncContext *s, PutBitContext *pb, const float *in, float *out, const float *scaled, int size, int scale_idx, int cb, const float lambda, const float uplim, int *bits, float *energy, int BT_ZERO, int BT_UNSIGNED, int BT_PAIR, int BT_ESC, int BT_NOISE, int BT_STEREO, const float ROUNDING)
Calculate rate distortion cost for quantizing with given codebook.
float ff_aac_pow2sf_tab[428]
Definition: aactab.c:35
#define SCALE_DIV_512
scalefactor difference that corresponds to scale difference in 512 times
Definition: aac.h:148
float ff_aac_pow34sf_tab[428]
Definition: aactab.c:36
const uint8_t *const ff_aac_spectral_bits[11]
Definition: aactab.c:422
void(* quant_bands)(int *out, const float *in, const float *scaled, int size, int is_signed, int maxval, const float Q34, const float rounding)
Definition: aacenc.h:414
static int quantize_band_cost_bits(struct AACEncContext *s, const float *in, const float *scaled, int size, int scale_idx, int cb, const float lambda, const float uplim, int *bits, float *energy, int rtz)
PutBitContext pb
Definition: aacenc.h:379
#define ROUND_STANDARD
Definition: aacenc_utils.h:36
#define av_log2
Definition: intmath.h:83
static const uint8_t aac_cb_range[12]
Definition: aacenctab.h:125
static av_always_inline float cbrtf(float x)
Definition: libm.h:61
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(const int16_t *) pi >> 8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(const int32_t *) pi >> 24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(const float *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(const float *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(const float *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(const double *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(const double *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(const double *) pi *(1U<< 31)))) #define SET_CONV_FUNC_GROUP(ofmt, ifmt) static void set_generic_function(AudioConvert *ac) { } void ff_audio_convert_free(AudioConvert **ac) { if(! *ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);} AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enum AVSampleFormat out_fmt, enum AVSampleFormat in_fmt, int channels, int sample_rate, int apply_map) { AudioConvert *ac;int in_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) return NULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method !=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt) > 2) { ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc) { av_free(ac);return NULL;} return ac;} in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar) { ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar ? ac->channels :1;} else if(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;else ac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);return ac;} int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in) { int use_generic=1;int len=in->nb_samples;int p;if(ac->dc) { av_log(ac->avr, AV_LOG_TRACE, "%d samples - audio_convert: %s to %s (dithered)\", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));return ff_convert_dither(ac-> in
static float quantize_and_encode_band_cost_NONE(struct AACEncContext *s, PutBitContext *pb, const float *in, float *quant, const float *scaled, int size, int scale_idx, int cb, const float lambda, const float uplim, int *bits, float *energy)
int dim
AAC encoder data.
const uint8_t * quant
#define SCALE_ONE_POS
scalefactor index that corresponds to scale=1.0
Definition: aac.h:149
#define STEREO
Definition: cook.c:61
AAC encoder utilities.
static double c[64]
static const uint8_t aac_cb_maxval[12]
Definition: aacenctab.h:126
const uint16_t *const ff_aac_spectral_codes[11]
Definition: aactab.c:417
int len
#define QUANTIZE_AND_ENCODE_BAND_COST_FUNC(NAME, BT_ZERO, BT_UNSIGNED, BT_PAIR, BT_ESC, BT_NOISE, BT_STEREO, ROUNDING)
static void quantize_and_encode_band(struct AACEncContext *s, PutBitContext *pb, const float *in, float *out, int size, int scale_idx, int cb, const float lambda, int rtz)
AAC encoder quantization misc reusable function templates.
#define POW_SF2_ZERO
ff_aac_pow2sf_tab index corresponding to pow(2, 0);
Definition: aac.h:154
FILE * out
Definition: movenc.c:54
#define av_always_inline
Definition: attributes.h:39
static float quantize_band_cost(struct AACEncContext *s, const float *in, const float *scaled, int size, int scale_idx, int cb, const float lambda, const float uplim, int *bits, float *energy, int rtz)
void(* abs_pow34)(float *out, const float *in, const int size)
Definition: aacenc.h:413
#define INFINITY
Definition: mathematics.h:67
AAC data declarations.
float scoefs[1024]
scaled coefficients
Definition: aacenc.h:408