72033b93102f2349f3363ff0b5413e1cea61cbe7
[libav.git] / libavcodec / libx264.c
1 /*
2 * H.264 encoding using the x264 library
3 * Copyright (C) 2005 Mans Rullgard <mans@mansr.com>
4 *
5 * This file is part of Libav.
6 *
7 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include "libavutil/opt.h"
23 #include "avcodec.h"
24 #include "internal.h"
25 #include <x264.h>
26 #include <float.h>
27 #include <math.h>
28 #include <stdio.h>
29 #include <stdlib.h>
30 #include <string.h>
31
32 typedef struct X264Context {
33 AVClass *class;
34 x264_param_t params;
35 x264_t *enc;
36 x264_picture_t pic;
37 uint8_t *sei;
38 int sei_size;
39 AVFrame out_pic;
40 char *preset;
41 char *tune;
42 char *profile;
43 int fastfirstpass;
44 float crf;
45 int cqp;
46 int aq_mode;
47 float aq_strength;
48 float psy_rd;
49 float psy_trellis;
50 int rc_lookahead;
51 } X264Context;
52
53 static void X264_log(void *p, int level, const char *fmt, va_list args)
54 {
55 static const int level_map[] = {
56 [X264_LOG_ERROR] = AV_LOG_ERROR,
57 [X264_LOG_WARNING] = AV_LOG_WARNING,
58 [X264_LOG_INFO] = AV_LOG_INFO,
59 [X264_LOG_DEBUG] = AV_LOG_DEBUG
60 };
61
62 if (level < 0 || level > X264_LOG_DEBUG)
63 return;
64
65 av_vlog(p, level_map[level], fmt, args);
66 }
67
68
69 static int encode_nals(AVCodecContext *ctx, uint8_t *buf, int size,
70 x264_nal_t *nals, int nnal, int skip_sei)
71 {
72 X264Context *x4 = ctx->priv_data;
73 uint8_t *p = buf;
74 int i;
75
76 /* Write the SEI as part of the first frame. */
77 if (x4->sei_size > 0 && nnal > 0) {
78 memcpy(p, x4->sei, x4->sei_size);
79 p += x4->sei_size;
80 x4->sei_size = 0;
81 }
82
83 for (i = 0; i < nnal; i++){
84 /* Don't put the SEI in extradata. */
85 if (skip_sei && nals[i].i_type == NAL_SEI) {
86 x4->sei_size = nals[i].i_payload;
87 x4->sei = av_malloc(x4->sei_size);
88 memcpy(x4->sei, nals[i].p_payload, nals[i].i_payload);
89 continue;
90 }
91 memcpy(p, nals[i].p_payload, nals[i].i_payload);
92 p += nals[i].i_payload;
93 }
94
95 return p - buf;
96 }
97
98 static int X264_frame(AVCodecContext *ctx, uint8_t *buf,
99 int bufsize, void *data)
100 {
101 X264Context *x4 = ctx->priv_data;
102 AVFrame *frame = data;
103 x264_nal_t *nal;
104 int nnal, i;
105 x264_picture_t pic_out;
106
107 x264_picture_init( &x4->pic );
108 x4->pic.img.i_csp = X264_CSP_I420;
109 x4->pic.img.i_plane = 3;
110
111 if (frame) {
112 for (i = 0; i < 3; i++) {
113 x4->pic.img.plane[i] = frame->data[i];
114 x4->pic.img.i_stride[i] = frame->linesize[i];
115 }
116
117 x4->pic.i_pts = frame->pts;
118 x4->pic.i_type =
119 frame->pict_type == AV_PICTURE_TYPE_I ? X264_TYPE_KEYFRAME :
120 frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
121 frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
122 X264_TYPE_AUTO;
123 if (x4->params.b_tff != frame->top_field_first) {
124 x4->params.b_tff = frame->top_field_first;
125 x264_encoder_reconfig(x4->enc, &x4->params);
126 }
127 }
128
129 do {
130 if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
131 return -1;
132
133 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
134 if (bufsize < 0)
135 return -1;
136 } while (!bufsize && !frame && x264_encoder_delayed_frames(x4->enc));
137
138 /* FIXME: libx264 now provides DTS, but AVFrame doesn't have a field for it. */
139 x4->out_pic.pts = pic_out.i_pts;
140
141 switch (pic_out.i_type) {
142 case X264_TYPE_IDR:
143 case X264_TYPE_I:
144 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
145 break;
146 case X264_TYPE_P:
147 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
148 break;
149 case X264_TYPE_B:
150 case X264_TYPE_BREF:
151 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
152 break;
153 }
154
155 x4->out_pic.key_frame = pic_out.b_keyframe;
156 if (bufsize)
157 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
158
159 return bufsize;
160 }
161
162 static av_cold int X264_close(AVCodecContext *avctx)
163 {
164 X264Context *x4 = avctx->priv_data;
165
166 av_freep(&avctx->extradata);
167 av_free(x4->sei);
168
169 if (x4->enc)
170 x264_encoder_close(x4->enc);
171
172 return 0;
173 }
174
175 static av_cold int X264_init(AVCodecContext *avctx)
176 {
177 X264Context *x4 = avctx->priv_data;
178
179 x4->sei_size = 0;
180 x264_param_default(&x4->params);
181
182 x4->params.i_keyint_max = avctx->gop_size;
183
184 x4->params.i_bframe = avctx->max_b_frames;
185 x4->params.b_cabac = avctx->coder_type == FF_CODER_TYPE_AC;
186 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
187 x4->params.i_bframe_bias = avctx->bframebias;
188 x4->params.i_bframe_pyramid = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? X264_B_PYRAMID_NORMAL : X264_B_PYRAMID_NONE;
189 avctx->has_b_frames = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? 2 : !!avctx->max_b_frames;
190
191 x4->params.i_keyint_min = avctx->keyint_min;
192 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
193 x4->params.i_keyint_min = x4->params.i_keyint_max;
194
195 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
196
197 x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
198 x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha;
199 x4->params.i_deblocking_filter_beta = avctx->deblockbeta;
200
201 x4->params.rc.i_qp_min = avctx->qmin;
202 x4->params.rc.i_qp_max = avctx->qmax;
203 x4->params.rc.i_qp_step = avctx->max_qdiff;
204
205 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
206 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
207 x4->params.rc.f_complexity_blur = avctx->complexityblur;
208
209 x4->params.i_frame_reference = avctx->refs;
210
211 x4->params.analyse.inter = 0;
212 if (avctx->partitions) {
213 if (avctx->partitions & X264_PART_I4X4)
214 x4->params.analyse.inter |= X264_ANALYSE_I4x4;
215 if (avctx->partitions & X264_PART_I8X8)
216 x4->params.analyse.inter |= X264_ANALYSE_I8x8;
217 if (avctx->partitions & X264_PART_P8X8)
218 x4->params.analyse.inter |= X264_ANALYSE_PSUB16x16;
219 if (avctx->partitions & X264_PART_P4X4)
220 x4->params.analyse.inter |= X264_ANALYSE_PSUB8x8;
221 if (avctx->partitions & X264_PART_B8X8)
222 x4->params.analyse.inter |= X264_ANALYSE_BSUB16x16;
223 }
224
225 x4->params.analyse.i_direct_mv_pred = avctx->directpred;
226
227 x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED;
228 x4->params.analyse.i_weighted_pred = avctx->weighted_p_pred;
229
230 if (avctx->me_method == ME_EPZS)
231 x4->params.analyse.i_me_method = X264_ME_DIA;
232 else if (avctx->me_method == ME_HEX)
233 x4->params.analyse.i_me_method = X264_ME_HEX;
234 else if (avctx->me_method == ME_UMH)
235 x4->params.analyse.i_me_method = X264_ME_UMH;
236 else if (avctx->me_method == ME_FULL)
237 x4->params.analyse.i_me_method = X264_ME_ESA;
238 else if (avctx->me_method == ME_TESA)
239 x4->params.analyse.i_me_method = X264_ME_TESA;
240 else x4->params.analyse.i_me_method = X264_ME_HEX;
241
242 x4->params.analyse.b_psy = avctx->flags2 & CODEC_FLAG2_PSY;
243
244 x4->params.analyse.i_me_range = avctx->me_range;
245 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
246
247 x4->params.analyse.b_mixed_references = avctx->flags2 & CODEC_FLAG2_MIXED_REFS;
248 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
249 x4->params.analyse.b_transform_8x8 = avctx->flags2 & CODEC_FLAG2_8X8DCT;
250 x4->params.analyse.b_fast_pskip = avctx->flags2 & CODEC_FLAG2_FASTPSKIP;
251
252 x4->params.analyse.i_trellis = avctx->trellis;
253 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
254
255 if (avctx->level > 0)
256 x4->params.i_level_idc = avctx->level;
257
258 if (x4->preset || x4->tune)
259 if (x264_param_default_preset(&x4->params, x4->preset, x4->tune) < 0) {
260 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", x4->preset, x4->tune);
261 return AVERROR(EINVAL);
262 }
263
264 x4->params.pf_log = X264_log;
265 x4->params.p_log_private = avctx;
266 x4->params.i_log_level = X264_LOG_DEBUG;
267
268 x4->params.b_intra_refresh = avctx->flags2 & CODEC_FLAG2_INTRA_REFRESH;
269 if (avctx->bit_rate) {
270 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
271 x4->params.rc.i_rc_method = X264_RC_ABR;
272 }
273 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
274 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
275 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
276 if (avctx->flags & CODEC_FLAG_PASS2) {
277 x4->params.rc.b_stat_read = 1;
278 } else {
279 #if FF_API_X264_GLOBAL_OPTS
280 if (avctx->crf) {
281 x4->params.rc.i_rc_method = X264_RC_CRF;
282 x4->params.rc.f_rf_constant = avctx->crf;
283 x4->params.rc.f_rf_constant_max = avctx->crf_max;
284 } else if (avctx->cqp > -1) {
285 x4->params.rc.i_rc_method = X264_RC_CQP;
286 x4->params.rc.i_qp_constant = avctx->cqp;
287 }
288 #endif
289
290 if (x4->crf >= 0) {
291 x4->params.rc.i_rc_method = X264_RC_CRF;
292 x4->params.rc.f_rf_constant = x4->crf;
293 } else if (x4->cqp >= 0) {
294 x4->params.rc.i_rc_method = X264_RC_CQP;
295 x4->params.rc.i_qp_constant = x4->cqp;
296 }
297 }
298
299 if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy &&
300 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
301 x4->params.rc.f_vbv_buffer_init =
302 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
303 }
304
305 x4->params.rc.b_mb_tree = !!(avctx->flags2 & CODEC_FLAG2_MBTREE);
306 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
307 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
308 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
309
310 #if FF_API_X264_GLOBAL_OPTS
311 if (avctx->aq_mode >= 0)
312 x4->params.rc.i_aq_mode = avctx->aq_mode;
313 if (avctx->aq_strength >= 0)
314 x4->params.rc.f_aq_strength = avctx->aq_strength;
315 if (avctx->psy_rd >= 0)
316 x4->params.analyse.f_psy_rd = avctx->psy_rd;
317 if (avctx->psy_trellis >= 0)
318 x4->params.analyse.f_psy_trellis = avctx->psy_trellis;
319 if (avctx->rc_lookahead >= 0)
320 x4->params.rc.i_lookahead = avctx->rc_lookahead;
321 #endif
322
323 if (x4->aq_mode >= 0)
324 x4->params.rc.i_aq_mode = x4->aq_mode;
325 if (x4->aq_strength >= 0)
326 x4->params.rc.f_aq_strength = x4->aq_strength;
327 if (x4->psy_rd >= 0)
328 x4->params.analyse.f_psy_rd = x4->psy_rd;
329 if (x4->psy_trellis >= 0)
330 x4->params.analyse.f_psy_trellis = x4->psy_trellis;
331 if (x4->rc_lookahead >= 0)
332 x4->params.rc.i_lookahead = x4->rc_lookahead;
333
334
335 if (x4->fastfirstpass)
336 x264_param_apply_fastfirstpass(&x4->params);
337
338 if (x4->profile)
339 if (x264_param_apply_profile(&x4->params, x4->profile) < 0) {
340 av_log(avctx, AV_LOG_ERROR, "Error setting profile %s.\n", x4->profile);
341 return AVERROR(EINVAL);
342 }
343
344 x4->params.i_width = avctx->width;
345 x4->params.i_height = avctx->height;
346 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
347 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
348 x4->params.i_fps_num = x4->params.i_timebase_den = avctx->time_base.den;
349 x4->params.i_fps_den = x4->params.i_timebase_num = avctx->time_base.num;
350
351 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
352 x4->params.analyse.b_ssim = avctx->flags2 & CODEC_FLAG2_SSIM;
353
354 x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD;
355
356 x4->params.i_threads = avctx->thread_count;
357
358 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
359
360 x4->params.b_open_gop = !(avctx->flags & CODEC_FLAG_CLOSED_GOP);
361
362 x4->params.i_slice_count = avctx->slices;
363
364 x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
365
366 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
367 x4->params.b_repeat_headers = 0;
368
369 // update AVCodecContext with x264 parameters
370 avctx->has_b_frames = x4->params.i_bframe ?
371 x4->params.i_bframe_pyramid ? 2 : 1 : 0;
372 avctx->bit_rate = x4->params.rc.i_bitrate*1000;
373 #if FF_API_X264_GLOBAL_OPTS
374 avctx->crf = x4->params.rc.f_rf_constant;
375 #endif
376
377 x4->enc = x264_encoder_open(&x4->params);
378 if (!x4->enc)
379 return -1;
380
381 avctx->coded_frame = &x4->out_pic;
382
383 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
384 x264_nal_t *nal;
385 int nnal, s, i;
386
387 s = x264_encoder_headers(x4->enc, &nal, &nnal);
388
389 for (i = 0; i < nnal; i++)
390 if (nal[i].i_type == NAL_SEI)
391 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
392
393 avctx->extradata = av_malloc(s);
394 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
395 }
396
397 return 0;
398 }
399
400 #define OFFSET(x) offsetof(X264Context, x)
401 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
402 static const AVOption options[] = {
403 { "preset", "Set the encoding preset (cf. x264 --fullhelp)", OFFSET(preset), FF_OPT_TYPE_STRING, { .str = "medium" }, 0, 0, VE},
404 { "tune", "Tune the encoding params (cf. x264 --fullhelp)", OFFSET(tune), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
405 { "profile", "Set profile restrictions (cf. x264 --fullhelp) ", OFFSET(profile), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
406 { "fastfirstpass", "Use fast settings when encoding first pass", OFFSET(fastfirstpass), FF_OPT_TYPE_INT, { 1 }, 0, 1, VE},
407 { "crf", "Select the quality for constant quality mode", OFFSET(crf), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
408 { "cqp", "Constant quantization parameter rate control method",OFFSET(cqp), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
409 { "aq_mode", "AQ method", OFFSET(aq_mode), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "aq_mode"},
410 { "none", NULL, 0, FF_OPT_TYPE_CONST, {X264_AQ_NONE}, INT_MIN, INT_MAX, VE, "aq_mode" },
411 { "variance", "Variance AQ (complexity mask)", 0, FF_OPT_TYPE_CONST, {X264_AQ_VARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
412 { "autovariance", "Auto-variance AQ (experimental)", 0, FF_OPT_TYPE_CONST, {X264_AQ_AUTOVARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
413 { "aq_strength", "AQ strength. Reduces blocking and blurring in flat and textured areas.", OFFSET(aq_strength), FF_OPT_TYPE_FLOAT, {-1}, -1, FLT_MAX, VE},
414 { "pdy_rd", "Psy RD strength.", OFFSET(psy_rd), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
415 { "psy_trellis", "Psy trellis strength", OFFSET(psy_trellis), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
416 { "rc_lookahead", "Number of frames to look ahead for frametype and ratecontrol", OFFSET(rc_lookahead), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
417 { NULL },
418 };
419
420 static const AVClass class = {
421 .class_name = "libx264",
422 .item_name = av_default_item_name,
423 .option = options,
424 .version = LIBAVUTIL_VERSION_INT,
425 };
426
427 static const AVCodecDefault x264_defaults[] = {
428 { "b", "0" },
429 { NULL },
430 };
431
432 AVCodec ff_libx264_encoder = {
433 .name = "libx264",
434 .type = AVMEDIA_TYPE_VIDEO,
435 .id = CODEC_ID_H264,
436 .priv_data_size = sizeof(X264Context),
437 .init = X264_init,
438 .encode = X264_frame,
439 .close = X264_close,
440 .capabilities = CODEC_CAP_DELAY,
441 .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_YUVJ420P, PIX_FMT_NONE },
442 .long_name = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
443 .priv_class = &class,
444 .defaults = x264_defaults,
445 };