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.

479 lines
17KB

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