cafca18451f46768e0248472b9955ceba877e58c
[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 } X264Context;
47
48 static void X264_log(void *p, int level, const char *fmt, va_list args)
49 {
50 static const int level_map[] = {
51 [X264_LOG_ERROR] = AV_LOG_ERROR,
52 [X264_LOG_WARNING] = AV_LOG_WARNING,
53 [X264_LOG_INFO] = AV_LOG_INFO,
54 [X264_LOG_DEBUG] = AV_LOG_DEBUG
55 };
56
57 if (level < 0 || level > X264_LOG_DEBUG)
58 return;
59
60 av_vlog(p, level_map[level], fmt, args);
61 }
62
63
64 static int encode_nals(AVCodecContext *ctx, uint8_t *buf, int size,
65 x264_nal_t *nals, int nnal, int skip_sei)
66 {
67 X264Context *x4 = ctx->priv_data;
68 uint8_t *p = buf;
69 int i;
70
71 /* Write the SEI as part of the first frame. */
72 if (x4->sei_size > 0 && nnal > 0) {
73 memcpy(p, x4->sei, x4->sei_size);
74 p += x4->sei_size;
75 x4->sei_size = 0;
76 }
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_size = nals[i].i_payload;
82 x4->sei = av_malloc(x4->sei_size);
83 memcpy(x4->sei, nals[i].p_payload, nals[i].i_payload);
84 continue;
85 }
86 memcpy(p, nals[i].p_payload, nals[i].i_payload);
87 p += nals[i].i_payload;
88 }
89
90 return p - buf;
91 }
92
93 static int X264_frame(AVCodecContext *ctx, uint8_t *buf,
94 int bufsize, void *data)
95 {
96 X264Context *x4 = ctx->priv_data;
97 AVFrame *frame = data;
98 x264_nal_t *nal;
99 int nnal, i;
100 x264_picture_t pic_out;
101
102 x264_picture_init( &x4->pic );
103 x4->pic.img.i_csp = X264_CSP_I420;
104 x4->pic.img.i_plane = 3;
105
106 if (frame) {
107 for (i = 0; i < 3; i++) {
108 x4->pic.img.plane[i] = frame->data[i];
109 x4->pic.img.i_stride[i] = frame->linesize[i];
110 }
111
112 x4->pic.i_pts = frame->pts;
113 x4->pic.i_type =
114 frame->pict_type == AV_PICTURE_TYPE_I ? X264_TYPE_KEYFRAME :
115 frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
116 frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
117 X264_TYPE_AUTO;
118 if (x4->params.b_tff != frame->top_field_first) {
119 x4->params.b_tff = frame->top_field_first;
120 x264_encoder_reconfig(x4->enc, &x4->params);
121 }
122 }
123
124 do {
125 if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
126 return -1;
127
128 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
129 if (bufsize < 0)
130 return -1;
131 } while (!bufsize && !frame && x264_encoder_delayed_frames(x4->enc));
132
133 /* FIXME: libx264 now provides DTS, but AVFrame doesn't have a field for it. */
134 x4->out_pic.pts = pic_out.i_pts;
135
136 switch (pic_out.i_type) {
137 case X264_TYPE_IDR:
138 case X264_TYPE_I:
139 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
140 break;
141 case X264_TYPE_P:
142 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
143 break;
144 case X264_TYPE_B:
145 case X264_TYPE_BREF:
146 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
147 break;
148 }
149
150 x4->out_pic.key_frame = pic_out.b_keyframe;
151 if (bufsize)
152 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
153
154 return bufsize;
155 }
156
157 static av_cold int X264_close(AVCodecContext *avctx)
158 {
159 X264Context *x4 = avctx->priv_data;
160
161 av_freep(&avctx->extradata);
162 av_free(x4->sei);
163
164 if (x4->enc)
165 x264_encoder_close(x4->enc);
166
167 return 0;
168 }
169
170 static av_cold int X264_init(AVCodecContext *avctx)
171 {
172 X264Context *x4 = avctx->priv_data;
173
174 x4->sei_size = 0;
175 x264_param_default(&x4->params);
176
177 x4->params.i_keyint_max = avctx->gop_size;
178
179 x4->params.i_bframe = avctx->max_b_frames;
180 x4->params.b_cabac = avctx->coder_type == FF_CODER_TYPE_AC;
181 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
182 x4->params.i_bframe_bias = avctx->bframebias;
183 x4->params.i_bframe_pyramid = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? X264_B_PYRAMID_NORMAL : X264_B_PYRAMID_NONE;
184 avctx->has_b_frames = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? 2 : !!avctx->max_b_frames;
185
186 x4->params.i_keyint_min = avctx->keyint_min;
187 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
188 x4->params.i_keyint_min = x4->params.i_keyint_max;
189
190 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
191
192 x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
193 x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha;
194 x4->params.i_deblocking_filter_beta = avctx->deblockbeta;
195
196 x4->params.rc.i_qp_min = avctx->qmin;
197 x4->params.rc.i_qp_max = avctx->qmax;
198 x4->params.rc.i_qp_step = avctx->max_qdiff;
199
200 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
201 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
202 x4->params.rc.f_complexity_blur = avctx->complexityblur;
203
204 x4->params.i_frame_reference = avctx->refs;
205
206 x4->params.analyse.inter = 0;
207 if (avctx->partitions) {
208 if (avctx->partitions & X264_PART_I4X4)
209 x4->params.analyse.inter |= X264_ANALYSE_I4x4;
210 if (avctx->partitions & X264_PART_I8X8)
211 x4->params.analyse.inter |= X264_ANALYSE_I8x8;
212 if (avctx->partitions & X264_PART_P8X8)
213 x4->params.analyse.inter |= X264_ANALYSE_PSUB16x16;
214 if (avctx->partitions & X264_PART_P4X4)
215 x4->params.analyse.inter |= X264_ANALYSE_PSUB8x8;
216 if (avctx->partitions & X264_PART_B8X8)
217 x4->params.analyse.inter |= X264_ANALYSE_BSUB16x16;
218 }
219
220 x4->params.analyse.i_direct_mv_pred = avctx->directpred;
221
222 x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED;
223 x4->params.analyse.i_weighted_pred = avctx->weighted_p_pred;
224
225 if (avctx->me_method == ME_EPZS)
226 x4->params.analyse.i_me_method = X264_ME_DIA;
227 else if (avctx->me_method == ME_HEX)
228 x4->params.analyse.i_me_method = X264_ME_HEX;
229 else if (avctx->me_method == ME_UMH)
230 x4->params.analyse.i_me_method = X264_ME_UMH;
231 else if (avctx->me_method == ME_FULL)
232 x4->params.analyse.i_me_method = X264_ME_ESA;
233 else if (avctx->me_method == ME_TESA)
234 x4->params.analyse.i_me_method = X264_ME_TESA;
235 else x4->params.analyse.i_me_method = X264_ME_HEX;
236
237 x4->params.rc.i_aq_mode = avctx->aq_mode;
238 x4->params.rc.f_aq_strength = avctx->aq_strength;
239 x4->params.rc.i_lookahead = avctx->rc_lookahead;
240
241 x4->params.analyse.b_psy = avctx->flags2 & CODEC_FLAG2_PSY;
242 x4->params.analyse.f_psy_rd = avctx->psy_rd;
243 x4->params.analyse.f_psy_trellis = avctx->psy_trellis;
244
245 x4->params.analyse.i_me_range = avctx->me_range;
246 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
247
248 x4->params.analyse.b_mixed_references = avctx->flags2 & CODEC_FLAG2_MIXED_REFS;
249 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
250 x4->params.analyse.b_transform_8x8 = avctx->flags2 & CODEC_FLAG2_8X8DCT;
251 x4->params.analyse.b_fast_pskip = avctx->flags2 & CODEC_FLAG2_FASTPSKIP;
252
253 x4->params.analyse.i_trellis = avctx->trellis;
254 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
255
256 if (avctx->level > 0)
257 x4->params.i_level_idc = avctx->level;
258
259 if (x4->preset || x4->tune)
260 if (x264_param_default_preset(&x4->params, x4->preset, x4->tune) < 0) {
261 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", x4->preset, x4->tune);
262 return AVERROR(EINVAL);
263 }
264
265 x4->params.pf_log = X264_log;
266 x4->params.p_log_private = avctx;
267 x4->params.i_log_level = X264_LOG_DEBUG;
268
269 x4->params.b_intra_refresh = avctx->flags2 & CODEC_FLAG2_INTRA_REFRESH;
270 if (avctx->bit_rate) {
271 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
272 x4->params.rc.i_rc_method = X264_RC_ABR;
273 }
274 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
275 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
276 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
277 if (avctx->flags & CODEC_FLAG_PASS2) {
278 x4->params.rc.b_stat_read = 1;
279 } else {
280 #if FF_API_X264_GLOBAL_OPTS
281 if (avctx->crf) {
282 x4->params.rc.i_rc_method = X264_RC_CRF;
283 x4->params.rc.f_rf_constant = avctx->crf;
284 x4->params.rc.f_rf_constant_max = avctx->crf_max;
285 } else if (avctx->cqp > -1) {
286 x4->params.rc.i_rc_method = X264_RC_CQP;
287 x4->params.rc.i_qp_constant = avctx->cqp;
288 }
289 #endif
290
291 if (x4->crf >= 0) {
292 x4->params.rc.i_rc_method = X264_RC_CRF;
293 x4->params.rc.f_rf_constant = x4->crf;
294 } else if (x4->cqp >= 0) {
295 x4->params.rc.i_rc_method = X264_RC_CQP;
296 x4->params.rc.i_qp_constant = x4->cqp;
297 }
298 }
299
300 if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy &&
301 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
302 x4->params.rc.f_vbv_buffer_init =
303 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
304 }
305
306 x4->params.rc.b_mb_tree = !!(avctx->flags2 & CODEC_FLAG2_MBTREE);
307 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
308 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
309 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
310
311 if (x4->fastfirstpass)
312 x264_param_apply_fastfirstpass(&x4->params);
313
314 if (x4->profile)
315 if (x264_param_apply_profile(&x4->params, x4->profile) < 0) {
316 av_log(avctx, AV_LOG_ERROR, "Error setting profile %s.\n", x4->profile);
317 return AVERROR(EINVAL);
318 }
319
320 x4->params.i_width = avctx->width;
321 x4->params.i_height = avctx->height;
322 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
323 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
324 x4->params.i_fps_num = x4->params.i_timebase_den = avctx->time_base.den;
325 x4->params.i_fps_den = x4->params.i_timebase_num = avctx->time_base.num;
326
327 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
328 x4->params.analyse.b_ssim = avctx->flags2 & CODEC_FLAG2_SSIM;
329
330 x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD;
331
332 x4->params.i_threads = avctx->thread_count;
333
334 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
335
336 x4->params.b_open_gop = !(avctx->flags & CODEC_FLAG_CLOSED_GOP);
337
338 x4->params.i_slice_count = avctx->slices;
339
340 x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
341
342 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
343 x4->params.b_repeat_headers = 0;
344
345 // update AVCodecContext with x264 parameters
346 avctx->has_b_frames = x4->params.i_bframe ?
347 x4->params.i_bframe_pyramid ? 2 : 1 : 0;
348 avctx->bit_rate = x4->params.rc.i_bitrate*1000;
349 #if FF_API_X264_GLOBAL_OPTS
350 avctx->crf = x4->params.rc.f_rf_constant;
351 #endif
352
353 x4->enc = x264_encoder_open(&x4->params);
354 if (!x4->enc)
355 return -1;
356
357 avctx->coded_frame = &x4->out_pic;
358
359 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
360 x264_nal_t *nal;
361 int nnal, s, i;
362
363 s = x264_encoder_headers(x4->enc, &nal, &nnal);
364
365 for (i = 0; i < nnal; i++)
366 if (nal[i].i_type == NAL_SEI)
367 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
368
369 avctx->extradata = av_malloc(s);
370 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
371 }
372
373 return 0;
374 }
375
376 #define OFFSET(x) offsetof(X264Context, x)
377 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
378 static const AVOption options[] = {
379 { "preset", "Set the encoding preset (cf. x264 --fullhelp)", OFFSET(preset), FF_OPT_TYPE_STRING, { .str = "medium" }, 0, 0, VE},
380 { "tune", "Tune the encoding params (cf. x264 --fullhelp)", OFFSET(tune), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
381 { "profile", "Set profile restrictions (cf. x264 --fullhelp) ", OFFSET(profile), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
382 { "fastfirstpass", "Use fast settings when encoding first pass", OFFSET(fastfirstpass), FF_OPT_TYPE_INT, { 1 }, 0, 1, VE},
383 { "crf", "Select the quality for constant quality mode", OFFSET(crf), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
384 { "cqp", "Constant quantization parameter rate control method",OFFSET(cqp), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
385 { NULL },
386 };
387
388 static const AVClass class = {
389 .class_name = "libx264",
390 .item_name = av_default_item_name,
391 .option = options,
392 .version = LIBAVUTIL_VERSION_INT,
393 };
394
395 static const AVCodecDefault x264_defaults[] = {
396 { "b", "0" },
397 { NULL },
398 };
399
400 AVCodec ff_libx264_encoder = {
401 .name = "libx264",
402 .type = AVMEDIA_TYPE_VIDEO,
403 .id = CODEC_ID_H264,
404 .priv_data_size = sizeof(X264Context),
405 .init = X264_init,
406 .encode = X264_frame,
407 .close = X264_close,
408 .capabilities = CODEC_CAP_DELAY,
409 .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_YUVJ420P, PIX_FMT_NONE },
410 .long_name = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
411 .priv_class = &class,
412 .defaults = x264_defaults,
413 };