/libavcodec/libxavs.c

http://github.com/FFmpeg/FFmpeg · C · 482 lines · 374 code · 74 blank · 34 comment · 56 complexity · 5f403360f51956aa9c33b34344ae14ac MD5 · raw file

  1. /*
  2. * AVS encoding using the xavs library
  3. * Copyright (C) 2010 Amanda, Y.N. Wu <amanda11192003@gmail.com>
  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. #include <stdio.h>
  22. #include <stdlib.h>
  23. #include <string.h>
  24. #include <math.h>
  25. #include <stdint.h>
  26. #include <float.h>
  27. #include <xavs.h>
  28. #include "avcodec.h"
  29. #include "internal.h"
  30. #include "libavutil/internal.h"
  31. #include "libavutil/mem.h"
  32. #include "libavutil/opt.h"
  33. #define END_OF_STREAM 0x001
  34. #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
  35. #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
  36. #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
  37. typedef struct XavsContext {
  38. AVClass *class;
  39. xavs_param_t params;
  40. xavs_t *enc;
  41. xavs_picture_t pic;
  42. uint8_t *sei;
  43. int sei_size;
  44. int end_of_stream;
  45. float crf;
  46. int cqp;
  47. int b_bias;
  48. float cplxblur;
  49. int direct_pred;
  50. int aud;
  51. int fast_pskip;
  52. int motion_est;
  53. int mbtree;
  54. int mixed_refs;
  55. int b_frame_strategy;
  56. int chroma_offset;
  57. int scenechange_threshold;
  58. int noise_reduction;
  59. int64_t *pts_buffer;
  60. int out_frame_count;
  61. } XavsContext;
  62. static void XAVS_log(void *p, int level, const char *fmt, va_list args)
  63. {
  64. static const int level_map[] = {
  65. [XAVS_LOG_ERROR] = AV_LOG_ERROR,
  66. [XAVS_LOG_WARNING] = AV_LOG_WARNING,
  67. [XAVS_LOG_INFO] = AV_LOG_INFO,
  68. [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
  69. };
  70. if (level < 0 || level > XAVS_LOG_DEBUG)
  71. return;
  72. av_vlog(p, level_map[level], fmt, args);
  73. }
  74. static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
  75. xavs_nal_t *nals, int nnal)
  76. {
  77. XavsContext *x4 = ctx->priv_data;
  78. uint8_t *p;
  79. int i, s, ret, size = x4->sei_size + AV_INPUT_BUFFER_MIN_SIZE;
  80. if (!nnal)
  81. return 0;
  82. for (i = 0; i < nnal; i++)
  83. size += nals[i].i_payload;
  84. if ((ret = ff_alloc_packet2(ctx, pkt, size, 0)) < 0)
  85. return ret;
  86. p = pkt->data;
  87. /* Write the SEI as part of the first frame. */
  88. if (x4->sei_size > 0 && nnal > 0) {
  89. memcpy(p, x4->sei, x4->sei_size);
  90. p += x4->sei_size;
  91. x4->sei_size = 0;
  92. }
  93. for (i = 0; i < nnal; i++) {
  94. s = xavs_nal_encode(p, &size, 1, nals + i);
  95. if (s < 0)
  96. return -1;
  97. p += s;
  98. }
  99. pkt->size = p - pkt->data;
  100. return 1;
  101. }
  102. static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
  103. const AVFrame *frame, int *got_packet)
  104. {
  105. XavsContext *x4 = avctx->priv_data;
  106. xavs_nal_t *nal;
  107. int nnal, i, ret;
  108. xavs_picture_t pic_out;
  109. int pict_type;
  110. x4->pic.img.i_csp = XAVS_CSP_I420;
  111. x4->pic.img.i_plane = 3;
  112. if (frame) {
  113. for (i = 0; i < 3; i++) {
  114. x4->pic.img.plane[i] = frame->data[i];
  115. x4->pic.img.i_stride[i] = frame->linesize[i];
  116. }
  117. x4->pic.i_pts = frame->pts;
  118. x4->pic.i_type = XAVS_TYPE_AUTO;
  119. x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
  120. }
  121. if (xavs_encoder_encode(x4->enc, &nal, &nnal,
  122. frame? &x4->pic: NULL, &pic_out) < 0)
  123. return -1;
  124. ret = encode_nals(avctx, pkt, nal, nnal);
  125. if (ret < 0)
  126. return -1;
  127. if (!ret) {
  128. if (!frame && !(x4->end_of_stream)) {
  129. if ((ret = ff_alloc_packet2(avctx, pkt, 4, 0)) < 0)
  130. return ret;
  131. pkt->data[0] = 0x0;
  132. pkt->data[1] = 0x0;
  133. pkt->data[2] = 0x01;
  134. pkt->data[3] = 0xb1;
  135. pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
  136. x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
  137. x4->end_of_stream = END_OF_STREAM;
  138. *got_packet = 1;
  139. }
  140. return 0;
  141. }
  142. #if FF_API_CODED_FRAME
  143. FF_DISABLE_DEPRECATION_WARNINGS
  144. avctx->coded_frame->pts = pic_out.i_pts;
  145. FF_ENABLE_DEPRECATION_WARNINGS
  146. #endif
  147. pkt->pts = pic_out.i_pts;
  148. if (avctx->has_b_frames) {
  149. if (!x4->out_frame_count)
  150. pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
  151. else
  152. pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
  153. } else
  154. pkt->dts = pkt->pts;
  155. switch (pic_out.i_type) {
  156. case XAVS_TYPE_IDR:
  157. case XAVS_TYPE_I:
  158. pict_type = AV_PICTURE_TYPE_I;
  159. break;
  160. case XAVS_TYPE_P:
  161. pict_type = AV_PICTURE_TYPE_P;
  162. break;
  163. case XAVS_TYPE_B:
  164. case XAVS_TYPE_BREF:
  165. pict_type = AV_PICTURE_TYPE_B;
  166. break;
  167. default:
  168. pict_type = AV_PICTURE_TYPE_NONE;
  169. }
  170. #if FF_API_CODED_FRAME
  171. FF_DISABLE_DEPRECATION_WARNINGS
  172. avctx->coded_frame->pict_type = pict_type;
  173. FF_ENABLE_DEPRECATION_WARNINGS
  174. #endif
  175. /* There is no IDR frame in AVS JiZhun */
  176. /* Sequence header is used as a flag */
  177. if (pic_out.i_type == XAVS_TYPE_I) {
  178. #if FF_API_CODED_FRAME
  179. FF_DISABLE_DEPRECATION_WARNINGS
  180. avctx->coded_frame->key_frame = 1;
  181. FF_ENABLE_DEPRECATION_WARNINGS
  182. #endif
  183. pkt->flags |= AV_PKT_FLAG_KEY;
  184. }
  185. #if FF_API_CODED_FRAME
  186. FF_DISABLE_DEPRECATION_WARNINGS
  187. avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
  188. FF_ENABLE_DEPRECATION_WARNINGS
  189. #endif
  190. ff_side_data_set_encoder_stats(pkt, (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA, NULL, 0, pict_type);
  191. x4->out_frame_count++;
  192. *got_packet = ret;
  193. return 0;
  194. }
  195. static av_cold int XAVS_close(AVCodecContext *avctx)
  196. {
  197. XavsContext *x4 = avctx->priv_data;
  198. av_freep(&avctx->extradata);
  199. av_freep(&x4->sei);
  200. av_freep(&x4->pts_buffer);
  201. if (x4->enc)
  202. xavs_encoder_close(x4->enc);
  203. return 0;
  204. }
  205. static av_cold int XAVS_init(AVCodecContext *avctx)
  206. {
  207. XavsContext *x4 = avctx->priv_data;
  208. x4->sei_size = 0;
  209. xavs_param_default(&x4->params);
  210. x4->params.pf_log = XAVS_log;
  211. x4->params.p_log_private = avctx;
  212. x4->params.i_keyint_max = avctx->gop_size;
  213. if (avctx->bit_rate) {
  214. x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
  215. x4->params.rc.i_rc_method = XAVS_RC_ABR;
  216. }
  217. x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
  218. x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
  219. x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
  220. if (avctx->flags & AV_CODEC_FLAG_PASS2) {
  221. x4->params.rc.b_stat_read = 1;
  222. } else {
  223. if (x4->crf >= 0) {
  224. x4->params.rc.i_rc_method = XAVS_RC_CRF;
  225. x4->params.rc.f_rf_constant = x4->crf;
  226. } else if (x4->cqp >= 0) {
  227. x4->params.rc.i_rc_method = XAVS_RC_CQP;
  228. x4->params.rc.i_qp_constant = x4->cqp;
  229. }
  230. }
  231. if (x4->aud >= 0)
  232. x4->params.b_aud = x4->aud;
  233. if (x4->mbtree >= 0)
  234. x4->params.rc.b_mb_tree = x4->mbtree;
  235. if (x4->direct_pred >= 0)
  236. x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
  237. if (x4->fast_pskip >= 0)
  238. x4->params.analyse.b_fast_pskip = x4->fast_pskip;
  239. if (x4->motion_est >= 0)
  240. x4->params.analyse.i_me_method = x4->motion_est;
  241. if (x4->mixed_refs >= 0)
  242. x4->params.analyse.b_mixed_references = x4->mixed_refs;
  243. if (x4->b_bias != INT_MIN)
  244. x4->params.i_bframe_bias = x4->b_bias;
  245. if (x4->cplxblur >= 0)
  246. x4->params.rc.f_complexity_blur = x4->cplxblur;
  247. x4->params.i_bframe = avctx->max_b_frames;
  248. /* cabac is not included in AVS JiZhun Profile */
  249. x4->params.b_cabac = 0;
  250. #if FF_API_PRIVATE_OPT
  251. FF_DISABLE_DEPRECATION_WARNINGS
  252. if (avctx->b_frame_strategy)
  253. x4->b_frame_strategy = avctx->b_frame_strategy;
  254. FF_ENABLE_DEPRECATION_WARNINGS
  255. #endif
  256. x4->params.i_bframe_adaptive = x4->b_frame_strategy;
  257. avctx->has_b_frames = !!avctx->max_b_frames;
  258. /* AVS doesn't allow B picture as reference */
  259. /* The max allowed reference frame number of B is 2 */
  260. x4->params.i_keyint_min = avctx->keyint_min;
  261. if (x4->params.i_keyint_min > x4->params.i_keyint_max)
  262. x4->params.i_keyint_min = x4->params.i_keyint_max;
  263. #if FF_API_PRIVATE_OPT
  264. FF_DISABLE_DEPRECATION_WARNINGS
  265. if (avctx->scenechange_threshold)
  266. x4->scenechange_threshold = avctx->scenechange_threshold;
  267. FF_ENABLE_DEPRECATION_WARNINGS
  268. #endif
  269. x4->params.i_scenecut_threshold = x4->scenechange_threshold;
  270. // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
  271. x4->params.rc.i_qp_min = avctx->qmin;
  272. x4->params.rc.i_qp_max = avctx->qmax;
  273. x4->params.rc.i_qp_step = avctx->max_qdiff;
  274. x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
  275. x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
  276. x4->params.i_frame_reference = avctx->refs;
  277. x4->params.i_width = avctx->width;
  278. x4->params.i_height = avctx->height;
  279. x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
  280. x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
  281. /* This is only used for counting the fps */
  282. x4->params.i_fps_num = avctx->time_base.den;
  283. x4->params.i_fps_den = avctx->time_base.num;
  284. x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
  285. x4->params.analyse.i_me_range = avctx->me_range;
  286. x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
  287. x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
  288. /* AVS P2 only enables 8x8 transform */
  289. x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
  290. x4->params.analyse.i_trellis = avctx->trellis;
  291. #if FF_API_PRIVATE_OPT
  292. FF_DISABLE_DEPRECATION_WARNINGS
  293. if (avctx->noise_reduction >= 0)
  294. x4->noise_reduction = avctx->noise_reduction;
  295. FF_ENABLE_DEPRECATION_WARNINGS
  296. #endif
  297. x4->params.analyse.i_noise_reduction = x4->noise_reduction;
  298. if (avctx->level > 0)
  299. x4->params.i_level_idc = avctx->level;
  300. if (avctx->bit_rate > 0)
  301. x4->params.rc.f_rate_tolerance =
  302. (float)avctx->bit_rate_tolerance / avctx->bit_rate;
  303. if ((avctx->rc_buffer_size) &&
  304. (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
  305. x4->params.rc.f_vbv_buffer_init =
  306. (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
  307. } else
  308. x4->params.rc.f_vbv_buffer_init = 0.9;
  309. /* TAG:do we have MB tree RC method */
  310. /* what is the RC method we are now using? Default NO */
  311. x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
  312. x4->params.rc.f_pb_factor = avctx->b_quant_factor;
  313. #if FF_API_PRIVATE_OPT
  314. FF_DISABLE_DEPRECATION_WARNINGS
  315. if (avctx->chromaoffset)
  316. x4->chroma_offset = avctx->chromaoffset;
  317. FF_ENABLE_DEPRECATION_WARNINGS
  318. #endif
  319. x4->params.analyse.i_chroma_qp_offset = x4->chroma_offset;
  320. x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
  321. x4->params.i_log_level = XAVS_LOG_DEBUG;
  322. x4->params.i_threads = avctx->thread_count;
  323. x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
  324. if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
  325. x4->params.b_repeat_headers = 0;
  326. x4->enc = xavs_encoder_open(&x4->params);
  327. if (!x4->enc)
  328. return -1;
  329. if (!(x4->pts_buffer = av_mallocz_array((avctx->max_b_frames+1), sizeof(*x4->pts_buffer))))
  330. return AVERROR(ENOMEM);
  331. /* TAG: Do we have GLOBAL HEADER in AVS */
  332. /* We Have PPS and SPS in AVS */
  333. if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER && 0) {
  334. xavs_nal_t *nal;
  335. int nnal, s, i, size;
  336. uint8_t *p;
  337. s = xavs_encoder_headers(x4->enc, &nal, &nnal);
  338. avctx->extradata = p = av_malloc(s);
  339. for (i = 0; i < nnal; i++) {
  340. /* Don't put the SEI in extradata. */
  341. if (nal[i].i_type == NAL_SEI) {
  342. x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
  343. if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
  344. return -1;
  345. continue;
  346. }
  347. size = xavs_nal_encode(p, &s, 1, nal + i);
  348. if (size < 0)
  349. return -1;
  350. p += size;
  351. }
  352. avctx->extradata_size = p - avctx->extradata;
  353. }
  354. return 0;
  355. }
  356. #define OFFSET(x) offsetof(XavsContext, x)
  357. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
  358. static const AVOption options[] = {
  359. { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE },
  360. { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
  361. { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
  362. { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE},
  363. { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
  364. { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
  365. { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
  366. { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
  367. { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
  368. { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
  369. { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
  370. { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE },
  371. { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
  372. { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = XAVS_ME_DIA }, -1, XAVS_ME_TESA, VE, "motion-est"},
  373. { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
  374. { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
  375. { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
  376. { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
  377. { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
  378. { "b_strategy", "Strategy to choose between I/P/B-frames", OFFSET(b_frame_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 2, VE},
  379. { "chromaoffset", "QP difference between chroma and luma", OFFSET(chroma_offset), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, VE},
  380. { "sc_threshold", "Scene change threshold", OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
  381. { "noise_reduction", "Noise reduction", OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
  382. { NULL },
  383. };
  384. static const AVClass xavs_class = {
  385. .class_name = "libxavs",
  386. .item_name = av_default_item_name,
  387. .option = options,
  388. .version = LIBAVUTIL_VERSION_INT,
  389. };
  390. static const AVCodecDefault xavs_defaults[] = {
  391. { "b", "0" },
  392. { NULL },
  393. };
  394. AVCodec ff_libxavs_encoder = {
  395. .name = "libxavs",
  396. .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
  397. .type = AVMEDIA_TYPE_VIDEO,
  398. .id = AV_CODEC_ID_CAVS,
  399. .priv_data_size = sizeof(XavsContext),
  400. .init = XAVS_init,
  401. .encode2 = XAVS_frame,
  402. .close = XAVS_close,
  403. .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
  404. .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
  405. .priv_class = &xavs_class,
  406. .defaults = xavs_defaults,
  407. .wrapper_name = "libxavs",
  408. };