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.

424 lines
15KB

  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/opt.h"
  31. #define END_OF_STREAM 0x001
  32. #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
  33. #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
  34. #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
  35. typedef struct XavsContext {
  36. xavs_param_t params;
  37. xavs_t *enc;
  38. xavs_picture_t pic;
  39. uint8_t *sei;
  40. int sei_size;
  41. AVFrame out_pic;
  42. int end_of_stream;
  43. float crf;
  44. int cqp;
  45. int b_bias;
  46. float cplxblur;
  47. int direct_pred;
  48. int aud;
  49. int fast_pskip;
  50. int mbtree;
  51. int mixed_refs;
  52. } XavsContext;
  53. static void XAVS_log(void *p, int level, const char *fmt, va_list args)
  54. {
  55. static const int level_map[] = {
  56. [XAVS_LOG_ERROR] = AV_LOG_ERROR,
  57. [XAVS_LOG_WARNING] = AV_LOG_WARNING,
  58. [XAVS_LOG_INFO] = AV_LOG_INFO,
  59. [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
  60. };
  61. if (level < 0 || level > XAVS_LOG_DEBUG)
  62. return;
  63. av_vlog(p, level_map[level], fmt, args);
  64. }
  65. static int encode_nals(AVCodecContext *ctx, uint8_t *buf,
  66. int size, xavs_nal_t *nals,
  67. int nnal, int skip_sei)
  68. {
  69. XavsContext *x4 = ctx->priv_data;
  70. uint8_t *p = buf;
  71. int i, s;
  72. /* Write the SEI as part of the first frame. */
  73. if (x4->sei_size > 0 && nnal > 0) {
  74. memcpy(p, x4->sei, x4->sei_size);
  75. p += x4->sei_size;
  76. x4->sei_size = 0;
  77. }
  78. for (i = 0; i < nnal; i++) {
  79. /* Don't put the SEI in extradata. */
  80. if (skip_sei && nals[i].i_type == NAL_SEI) {
  81. x4->sei = av_malloc( 5 + nals[i].i_payload * 4 / 3 );
  82. if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nals + i) < 0)
  83. return -1;
  84. continue;
  85. }
  86. s = xavs_nal_encode(p, &size, 1, nals + i);
  87. if (s < 0)
  88. return -1;
  89. p += s;
  90. }
  91. return p - buf;
  92. }
  93. static int XAVS_frame(AVCodecContext *ctx, uint8_t *buf,
  94. int bufsize, void *data)
  95. {
  96. XavsContext *x4 = ctx->priv_data;
  97. AVFrame *frame = data;
  98. xavs_nal_t *nal;
  99. int nnal, i;
  100. xavs_picture_t pic_out;
  101. x4->pic.img.i_csp = XAVS_CSP_I420;
  102. x4->pic.img.i_plane = 3;
  103. if (frame) {
  104. for (i = 0; i < 3; i++) {
  105. x4->pic.img.plane[i] = frame->data[i];
  106. x4->pic.img.i_stride[i] = frame->linesize[i];
  107. }
  108. x4->pic.i_pts = frame->pts;
  109. x4->pic.i_type = XAVS_TYPE_AUTO;
  110. }
  111. if (xavs_encoder_encode(x4->enc, &nal, &nnal,
  112. frame? &x4->pic: NULL, &pic_out) < 0)
  113. return -1;
  114. bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
  115. if (bufsize < 0)
  116. return -1;
  117. if (!bufsize && !frame && !(x4->end_of_stream)){
  118. buf[bufsize] = 0x0;
  119. buf[bufsize+1] = 0x0;
  120. buf[bufsize+2] = 0x01;
  121. buf[bufsize+3] = 0xb1;
  122. bufsize += 4;
  123. x4->end_of_stream = END_OF_STREAM;
  124. return bufsize;
  125. }
  126. /* FIXME: libxavs now provides DTS */
  127. /* but AVFrame doesn't have a field for it. */
  128. x4->out_pic.pts = pic_out.i_pts;
  129. switch (pic_out.i_type) {
  130. case XAVS_TYPE_IDR:
  131. case XAVS_TYPE_I:
  132. x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
  133. break;
  134. case XAVS_TYPE_P:
  135. x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
  136. break;
  137. case XAVS_TYPE_B:
  138. case XAVS_TYPE_BREF:
  139. x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
  140. break;
  141. }
  142. /* There is no IDR frame in AVS JiZhun */
  143. /* Sequence header is used as a flag */
  144. x4->out_pic.key_frame = pic_out.i_type == XAVS_TYPE_I;
  145. x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
  146. return bufsize;
  147. }
  148. static av_cold int XAVS_close(AVCodecContext *avctx)
  149. {
  150. XavsContext *x4 = avctx->priv_data;
  151. av_freep(&avctx->extradata);
  152. av_free(x4->sei);
  153. if (x4->enc)
  154. xavs_encoder_close(x4->enc);
  155. return 0;
  156. }
  157. static av_cold int XAVS_init(AVCodecContext *avctx)
  158. {
  159. XavsContext *x4 = avctx->priv_data;
  160. x4->sei_size = 0;
  161. xavs_param_default(&x4->params);
  162. x4->params.pf_log = XAVS_log;
  163. x4->params.p_log_private = avctx;
  164. x4->params.i_keyint_max = avctx->gop_size;
  165. if (avctx->bit_rate) {
  166. x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
  167. x4->params.rc.i_rc_method = XAVS_RC_ABR;
  168. }
  169. x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
  170. x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
  171. x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
  172. if (avctx->flags & CODEC_FLAG_PASS2) {
  173. x4->params.rc.b_stat_read = 1;
  174. } else {
  175. #if FF_API_X264_GLOBAL_OPTS
  176. if (avctx->crf) {
  177. x4->params.rc.i_rc_method = XAVS_RC_CRF;
  178. x4->params.rc.f_rf_constant = avctx->crf;
  179. } else if (avctx->cqp > -1) {
  180. x4->params.rc.i_rc_method = XAVS_RC_CQP;
  181. x4->params.rc.i_qp_constant = avctx->cqp;
  182. }
  183. #endif
  184. if (x4->crf >= 0) {
  185. x4->params.rc.i_rc_method = XAVS_RC_CRF;
  186. x4->params.rc.f_rf_constant = x4->crf;
  187. } else if (x4->cqp >= 0) {
  188. x4->params.rc.i_rc_method = XAVS_RC_CQP;
  189. x4->params.rc.i_qp_constant = x4->cqp;
  190. }
  191. }
  192. #if FF_API_X264_GLOBAL_OPTS
  193. if (avctx->bframebias)
  194. x4->params.i_bframe_bias = avctx->bframebias;
  195. if (avctx->deblockalpha)
  196. x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha;
  197. if (avctx->deblockbeta)
  198. x4->params.i_deblocking_filter_beta = avctx->deblockbeta;
  199. if (avctx->complexityblur >= 0)
  200. x4->params.rc.f_complexity_blur = avctx->complexityblur;
  201. if (avctx->directpred >= 0)
  202. x4->params.analyse.i_direct_mv_pred = avctx->directpred;
  203. if (avctx->partitions) {
  204. if (avctx->partitions & XAVS_PART_I8X8)
  205. x4->params.analyse.inter |= XAVS_ANALYSE_I8x8;
  206. if (avctx->partitions & XAVS_PART_P8X8)
  207. x4->params.analyse.inter |= XAVS_ANALYSE_PSUB16x16;
  208. if (avctx->partitions & XAVS_PART_B8X8)
  209. x4->params.analyse.inter |= XAVS_ANALYSE_BSUB16x16;
  210. }
  211. x4->params.rc.b_mb_tree = !!(avctx->flags2 & CODEC_FLAG2_MBTREE);
  212. x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD;
  213. x4->params.analyse.b_mixed_references = avctx->flags2 & CODEC_FLAG2_MIXED_REFS;
  214. x4->params.analyse.b_fast_pskip = avctx->flags2 & CODEC_FLAG2_FASTPSKIP;
  215. x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED;
  216. #endif
  217. if (x4->aud >= 0)
  218. x4->params.b_aud = x4->aud;
  219. if (x4->mbtree >= 0)
  220. x4->params.rc.b_mb_tree = x4->mbtree;
  221. if (x4->direct_pred >= 0)
  222. x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
  223. if (x4->fast_pskip >= 0)
  224. x4->params.analyse.b_fast_pskip = x4->fast_pskip;
  225. if (x4->mixed_refs >= 0)
  226. x4->params.analyse.b_mixed_references = x4->mixed_refs;
  227. if (x4->b_bias != INT_MIN)
  228. x4->params.i_bframe_bias = x4->b_bias;
  229. if (x4->cplxblur >= 0)
  230. x4->params.rc.f_complexity_blur = x4->cplxblur;
  231. x4->params.i_bframe = avctx->max_b_frames;
  232. /* cabac is not included in AVS JiZhun Profile */
  233. x4->params.b_cabac = 0;
  234. x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
  235. avctx->has_b_frames = !!avctx->max_b_frames;
  236. /* AVS doesn't allow B picture as reference */
  237. /* The max allowed reference frame number of B is 2 */
  238. x4->params.i_keyint_min = avctx->keyint_min;
  239. if (x4->params.i_keyint_min > x4->params.i_keyint_max)
  240. x4->params.i_keyint_min = x4->params.i_keyint_max;
  241. x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
  242. // x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
  243. x4->params.rc.i_qp_min = avctx->qmin;
  244. x4->params.rc.i_qp_max = avctx->qmax;
  245. x4->params.rc.i_qp_step = avctx->max_qdiff;
  246. x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
  247. x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
  248. x4->params.i_frame_reference = avctx->refs;
  249. x4->params.i_width = avctx->width;
  250. x4->params.i_height = avctx->height;
  251. x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
  252. x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
  253. /* This is only used for counting the fps */
  254. x4->params.i_fps_num = avctx->time_base.den;
  255. x4->params.i_fps_den = avctx->time_base.num;
  256. x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
  257. switch (avctx->me_method) {
  258. case ME_EPZS:
  259. x4->params.analyse.i_me_method = XAVS_ME_DIA;
  260. break;
  261. case ME_HEX:
  262. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  263. break;
  264. case ME_UMH:
  265. x4->params.analyse.i_me_method = XAVS_ME_UMH;
  266. break;
  267. case ME_FULL:
  268. x4->params.analyse.i_me_method = XAVS_ME_ESA;
  269. break;
  270. case ME_TESA:
  271. x4->params.analyse.i_me_method = XAVS_ME_TESA;
  272. break;
  273. default:
  274. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  275. }
  276. x4->params.analyse.i_me_range = avctx->me_range;
  277. x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
  278. x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
  279. /* AVS P2 only enables 8x8 transform */
  280. x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & CODEC_FLAG2_8X8DCT;
  281. x4->params.analyse.i_trellis = avctx->trellis;
  282. x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
  283. if (avctx->level > 0)
  284. x4->params.i_level_idc = avctx->level;
  285. x4->params.rc.f_rate_tolerance =
  286. (float)avctx->bit_rate_tolerance/avctx->bit_rate;
  287. if ((avctx->rc_buffer_size) &&
  288. (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
  289. x4->params.rc.f_vbv_buffer_init =
  290. (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
  291. } else
  292. x4->params.rc.f_vbv_buffer_init = 0.9;
  293. /* TAG:do we have MB tree RC method */
  294. /* what is the RC method we are now using? Default NO */
  295. x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
  296. x4->params.rc.f_pb_factor = avctx->b_quant_factor;
  297. x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
  298. x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
  299. x4->params.i_log_level = XAVS_LOG_DEBUG;
  300. x4->params.i_threads = avctx->thread_count;
  301. x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
  302. if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
  303. x4->params.b_repeat_headers = 0;
  304. x4->enc = xavs_encoder_open(&x4->params);
  305. if (!x4->enc)
  306. return -1;
  307. avctx->coded_frame = &x4->out_pic;
  308. /* TAG: Do we have GLOBAL HEADER in AVS */
  309. /* We Have PPS and SPS in AVS */
  310. if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
  311. xavs_nal_t *nal;
  312. int nnal, s;
  313. s = xavs_encoder_headers(x4->enc, &nal, &nnal);
  314. avctx->extradata = av_malloc(s);
  315. avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
  316. }
  317. return 0;
  318. }
  319. #define OFFSET(x) offsetof(XavsContext, x)
  320. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
  321. static const AVOption options[] = {
  322. { "crf", "Select the quality for constant quality mode", OFFSET(crf), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
  323. { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
  324. { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), FF_OPT_TYPE_INT, {INT_MIN}, INT_MIN, INT_MAX, VE },
  325. { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
  326. { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "direct-pred" },
  327. { "none", NULL, 0, FF_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
  328. { "spatial", NULL, 0, FF_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
  329. { "temporal", NULL, 0, FF_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
  330. { "auto", NULL, 0, FF_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
  331. { "aud", "Use access unit delimiters.", OFFSET(aud), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE},
  332. { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE},
  333. { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), FF_OPT_TYPE_INT, {-1}, -1, 1, VE },
  334. { "fast-pskip", NULL, OFFSET(fast_pskip), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE},
  335. { NULL },
  336. };
  337. static const AVClass class = {
  338. .class_name = "libxavs",
  339. .item_name = av_default_item_name,
  340. .option = options,
  341. .version = LIBAVUTIL_VERSION_INT,
  342. };
  343. static const AVCodecDefault xavs_defaults[] = {
  344. { "b", "0" },
  345. { NULL },
  346. };
  347. AVCodec ff_libxavs_encoder = {
  348. .name = "libxavs",
  349. .type = AVMEDIA_TYPE_VIDEO,
  350. .id = CODEC_ID_CAVS,
  351. .priv_data_size = sizeof(XavsContext),
  352. .init = XAVS_init,
  353. .encode = XAVS_frame,
  354. .close = XAVS_close,
  355. .capabilities = CODEC_CAP_DELAY,
  356. .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_NONE },
  357. .long_name = NULL_IF_CONFIG_SMALL("libxavs - the Chinese Audio Video Standard Encoder"),
  358. .priv_class = &class,
  359. .defaults = xavs_defaults,
  360. };