You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

244 lines
8.8KB

  1. /*
  2. * AAC encoder utilities
  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. * @file
  23. * AAC encoder utilities
  24. * @author Rostislav Pehlivanov ( atomnuker gmail com )
  25. */
  26. #ifndef AVCODEC_AACENC_UTILS_H
  27. #define AVCODEC_AACENC_UTILS_H
  28. #include "aac.h"
  29. #include "aac_tablegen_decl.h"
  30. #include "aacenctab.h"
  31. #define ROUND_STANDARD 0.4054f
  32. #define ROUND_TO_ZERO 0.1054f
  33. #define C_QUANT 0.4054f
  34. static inline void abs_pow34_v(float *out, const float *in, const int size)
  35. {
  36. int i;
  37. for (i = 0; i < size; i++) {
  38. float a = fabsf(in[i]);
  39. out[i] = sqrtf(a * sqrtf(a));
  40. }
  41. }
  42. /**
  43. * Quantize one coefficient.
  44. * @return absolute value of the quantized coefficient
  45. * @see 3GPP TS26.403 5.6.2 "Scalefactor determination"
  46. */
  47. static inline int quant(float coef, const float Q, const float rounding)
  48. {
  49. float a = coef * Q;
  50. return sqrtf(a * sqrtf(a)) + rounding;
  51. }
  52. static inline void quantize_bands(int *out, const float *in, const float *scaled,
  53. int size, float Q34, int is_signed, int maxval,
  54. const float rounding)
  55. {
  56. int i;
  57. double qc;
  58. for (i = 0; i < size; i++) {
  59. qc = scaled[i] * Q34;
  60. out[i] = (int)FFMIN(qc + rounding, (double)maxval);
  61. if (is_signed && in[i] < 0.0f) {
  62. out[i] = -out[i];
  63. }
  64. }
  65. }
  66. static inline float find_max_val(int group_len, int swb_size, const float *scaled)
  67. {
  68. float maxval = 0.0f;
  69. int w2, i;
  70. for (w2 = 0; w2 < group_len; w2++) {
  71. for (i = 0; i < swb_size; i++) {
  72. maxval = FFMAX(maxval, scaled[w2*128+i]);
  73. }
  74. }
  75. return maxval;
  76. }
  77. static inline int find_min_book(float maxval, int sf)
  78. {
  79. float Q = ff_aac_pow2sf_tab[POW_SF2_ZERO - sf + SCALE_ONE_POS - SCALE_DIV_512];
  80. float Q34 = sqrtf(Q * sqrtf(Q));
  81. int qmaxval, cb;
  82. qmaxval = maxval * Q34 + C_QUANT;
  83. if (qmaxval >= (FF_ARRAY_ELEMS(aac_maxval_cb)))
  84. cb = 11;
  85. else
  86. cb = aac_maxval_cb[qmaxval];
  87. return cb;
  88. }
  89. static inline float find_form_factor(int group_len, int swb_size, float thresh, const float *scaled, float nzslope) {
  90. const float iswb_size = 1.0f / swb_size;
  91. const float iswb_sizem1 = 1.0f / (swb_size - 1);
  92. const float ethresh = thresh;
  93. float form = 0.0f, weight = 0.0f;
  94. int w2, i;
  95. for (w2 = 0; w2 < group_len; w2++) {
  96. float e = 0.0f, e2 = 0.0f, var = 0.0f, maxval = 0.0f;
  97. float nzl = 0;
  98. for (i = 0; i < swb_size; i++) {
  99. float s = fabsf(scaled[w2*128+i]);
  100. maxval = FFMAX(maxval, s);
  101. e += s;
  102. e2 += s *= s;
  103. /* We really don't want a hard non-zero-line count, since
  104. * even below-threshold lines do add up towards band spectral power.
  105. * So, fall steeply towards zero, but smoothly
  106. */
  107. if (s >= ethresh) {
  108. nzl += 1.0f;
  109. } else {
  110. nzl += powf(s / ethresh, nzslope);
  111. }
  112. }
  113. if (e2 > thresh) {
  114. float frm;
  115. e *= iswb_size;
  116. /** compute variance */
  117. for (i = 0; i < swb_size; i++) {
  118. float d = fabsf(scaled[w2*128+i]) - e;
  119. var += d*d;
  120. }
  121. var = sqrtf(var * iswb_sizem1);
  122. e2 *= iswb_size;
  123. frm = e / FFMIN(e+4*var,maxval);
  124. form += e2 * sqrtf(frm) / FFMAX(0.5f,nzl);
  125. weight += e2;
  126. }
  127. }
  128. if (weight > 0) {
  129. return form / weight;
  130. } else {
  131. return 1.0f;
  132. }
  133. }
  134. /** Return the minimum scalefactor where the quantized coef does not clip. */
  135. static inline uint8_t coef2minsf(float coef)
  136. {
  137. return av_clip_uint8(log2f(coef)*4 - 69 + SCALE_ONE_POS - SCALE_DIV_512);
  138. }
  139. /** Return the maximum scalefactor where the quantized coef is not zero. */
  140. static inline uint8_t coef2maxsf(float coef)
  141. {
  142. return av_clip_uint8(log2f(coef)*4 + 6 + SCALE_ONE_POS - SCALE_DIV_512);
  143. }
  144. /*
  145. * Returns the closest possible index to an array of float values, given a value.
  146. */
  147. static inline int quant_array_idx(const float val, const float *arr, const int num)
  148. {
  149. int i, index = 0;
  150. float quant_min_err = INFINITY;
  151. for (i = 0; i < num; i++) {
  152. float error = (val - arr[i])*(val - arr[i]);
  153. if (error < quant_min_err) {
  154. quant_min_err = error;
  155. index = i;
  156. }
  157. }
  158. return index;
  159. }
  160. /**
  161. * approximates exp10f(-3.0f*(0.5f + 0.5f * cosf(FFMIN(b,15.5f) / 15.5f)))
  162. */
  163. static av_always_inline float bval2bmax(float b)
  164. {
  165. return 0.001f + 0.0035f * (b*b*b) / (15.5f*15.5f*15.5f);
  166. }
  167. /*
  168. * linear congruential pseudorandom number generator, copied from the decoder
  169. */
  170. static inline int lcg_random(unsigned previous_val)
  171. {
  172. union { unsigned u; int s; } v = { previous_val * 1664525u + 1013904223 };
  173. return v.s;
  174. }
  175. #define ERROR_IF(cond, ...) \
  176. if (cond) { \
  177. av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \
  178. return AVERROR(EINVAL); \
  179. }
  180. #define WARN_IF(cond, ...) \
  181. if (cond) { \
  182. av_log(avctx, AV_LOG_WARNING, __VA_ARGS__); \
  183. }
  184. #define AAC_OPT_SET(e_opt, p_opt, bypass, name) \
  185. ERROR_IF ((e_opt)->name == 1 && (p_opt)->name == OPT_BANNED, \
  186. "Profile %i does not allow %s\n", avctx->profile, #name); \
  187. ERROR_IF ((e_opt)->name == 0 && (p_opt)->name == OPT_REQUIRED, \
  188. "Option %s is a requirement for this profile (%i)\n", \
  189. #name, avctx->profile); \
  190. if ((e_opt)->name == 1 && (p_opt)->name == OPT_NEEDS_MAIN && \
  191. avctx->profile == FF_PROFILE_AAC_LOW) { \
  192. WARN_IF(1, "Profile %i does not allow for %s, setting profile to " \
  193. "\"aac_main\"(%i)\n", avctx->profile, #name, \
  194. FF_PROFILE_AAC_MAIN); \
  195. avctx->profile = FF_PROFILE_AAC_MAIN; \
  196. p_opt = &aacenc_profiles[FF_PROFILE_AAC_MAIN].opts; \
  197. } \
  198. if ((e_opt)->name == 1 && (p_opt)->name == OPT_NEEDS_LTP && \
  199. avctx->profile == FF_PROFILE_AAC_LOW) { \
  200. WARN_IF(1, "Profile %i does not allow for %s, setting profile to " \
  201. "\"aac_ltp\"(%i)\n", avctx->profile, #name, \
  202. FF_PROFILE_AAC_LTP); \
  203. avctx->profile = FF_PROFILE_AAC_LTP; \
  204. p_opt = &aacenc_profiles[FF_PROFILE_AAC_LTP].opts; \
  205. } \
  206. if ((e_opt)->name == OPT_AUTO) { \
  207. if ((p_opt)->name == OPT_BANNED) { \
  208. (e_opt)->name = 0; \
  209. } else if ((p_opt)->name == OPT_NEEDS_LTP) { \
  210. (e_opt)->name = 0; \
  211. } else if ((p_opt)->name == OPT_NEEDS_MAIN) { \
  212. (e_opt)->name = 0; \
  213. } else if ((p_opt)->name == OPT_REQUIRED) { \
  214. (e_opt)->name = 1; \
  215. } else if (bypass) { \
  216. (e_opt)->name = (e_opt)->name; \
  217. } else { \
  218. (e_opt)->name = (p_opt)->name; \
  219. } \
  220. } \
  221. av_log(avctx, AV_LOG_VERBOSE, "Option %s set to %i\n", #name, (e_opt)->name);
  222. #endif /* AVCODEC_AACENC_UTILS_H */