libx264: add 'weightp' private option.
[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 int weightp;
52 } X264Context;
53
54 static void X264_log(void *p, int level, const char *fmt, va_list args)
55 {
56 static const int level_map[] = {
57 [X264_LOG_ERROR] = AV_LOG_ERROR,
58 [X264_LOG_WARNING] = AV_LOG_WARNING,
59 [X264_LOG_INFO] = AV_LOG_INFO,
60 [X264_LOG_DEBUG] = AV_LOG_DEBUG
61 };
62
63 if (level < 0 || level > X264_LOG_DEBUG)
64 return;
65
66 av_vlog(p, level_map[level], fmt, args);
67 }
68
69
70 static int encode_nals(AVCodecContext *ctx, uint8_t *buf, int size,
71 x264_nal_t *nals, int nnal, int skip_sei)
72 {
73 X264Context *x4 = ctx->priv_data;
74 uint8_t *p = buf;
75 int i;
76
77 /* Write the SEI as part of the first frame. */
78 if (x4->sei_size > 0 && nnal > 0) {
79 memcpy(p, x4->sei, x4->sei_size);
80 p += x4->sei_size;
81 x4->sei_size = 0;
82 }
83
84 for (i = 0; i < nnal; i++){
85 /* Don't put the SEI in extradata. */
86 if (skip_sei && nals[i].i_type == NAL_SEI) {
87 x4->sei_size = nals[i].i_payload;
88 x4->sei = av_malloc(x4->sei_size);
89 memcpy(x4->sei, nals[i].p_payload, nals[i].i_payload);
90 continue;
91 }
92 memcpy(p, nals[i].p_payload, nals[i].i_payload);
93 p += nals[i].i_payload;
94 }
95
96 return p - buf;
97 }
98
99 static int X264_frame(AVCodecContext *ctx, uint8_t *buf,
100 int bufsize, void *data)
101 {
102 X264Context *x4 = ctx->priv_data;
103 AVFrame *frame = data;
104 x264_nal_t *nal;
105 int nnal, i;
106 x264_picture_t pic_out;
107
108 x264_picture_init( &x4->pic );
109 x4->pic.img.i_csp = X264_CSP_I420;
110 x4->pic.img.i_plane = 3;
111
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
118 x4->pic.i_pts = frame->pts;
119 x4->pic.i_type =
120 frame->pict_type == AV_PICTURE_TYPE_I ? X264_TYPE_KEYFRAME :
121 frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
122 frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
123 X264_TYPE_AUTO;
124 if (x4->params.b_tff != frame->top_field_first) {
125 x4->params.b_tff = frame->top_field_first;
126 x264_encoder_reconfig(x4->enc, &x4->params);
127 }
128 }
129
130 do {
131 if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
132 return -1;
133
134 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
135 if (bufsize < 0)
136 return -1;
137 } while (!bufsize && !frame && x264_encoder_delayed_frames(x4->enc));
138
139 /* FIXME: libx264 now provides DTS, but AVFrame doesn't have a field for it. */
140 x4->out_pic.pts = pic_out.i_pts;
141
142 switch (pic_out.i_type) {
143 case X264_TYPE_IDR:
144 case X264_TYPE_I:
145 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
146 break;
147 case X264_TYPE_P:
148 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
149 break;
150 case X264_TYPE_B:
151 case X264_TYPE_BREF:
152 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
153 break;
154 }
155
156 x4->out_pic.key_frame = pic_out.b_keyframe;
157 if (bufsize)
158 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
159
160 return bufsize;
161 }
162
163 static av_cold int X264_close(AVCodecContext *avctx)
164 {
165 X264Context *x4 = avctx->priv_data;
166
167 av_freep(&avctx->extradata);
168 av_free(x4->sei);
169
170 if (x4->enc)
171 x264_encoder_close(x4->enc);
172
173 return 0;
174 }
175
176 static av_cold int X264_init(AVCodecContext *avctx)
177 {
178 X264Context *x4 = avctx->priv_data;
179
180 x4->sei_size = 0;
181 x264_param_default(&x4->params);
182
183 x4->params.i_keyint_max = avctx->gop_size;
184
185 x4->params.i_bframe = avctx->max_b_frames;
186 x4->params.b_cabac = avctx->coder_type == FF_CODER_TYPE_AC;
187 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
188 x4->params.i_bframe_bias = avctx->bframebias;
189 x4->params.i_bframe_pyramid = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? X264_B_PYRAMID_NORMAL : X264_B_PYRAMID_NONE;
190 avctx->has_b_frames = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? 2 : !!avctx->max_b_frames;
191
192 x4->params.i_keyint_min = avctx->keyint_min;
193 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
194 x4->params.i_keyint_min = x4->params.i_keyint_max;
195
196 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
197
198 x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
199 x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha;
200 x4->params.i_deblocking_filter_beta = avctx->deblockbeta;
201
202 x4->params.rc.i_qp_min = avctx->qmin;
203 x4->params.rc.i_qp_max = avctx->qmax;
204 x4->params.rc.i_qp_step = avctx->max_qdiff;
205
206 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
207 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
208 x4->params.rc.f_complexity_blur = avctx->complexityblur;
209
210 x4->params.i_frame_reference = avctx->refs;
211
212 x4->params.analyse.inter = 0;
213 if (avctx->partitions) {
214 if (avctx->partitions & X264_PART_I4X4)
215 x4->params.analyse.inter |= X264_ANALYSE_I4x4;
216 if (avctx->partitions & X264_PART_I8X8)
217 x4->params.analyse.inter |= X264_ANALYSE_I8x8;
218 if (avctx->partitions & X264_PART_P8X8)
219 x4->params.analyse.inter |= X264_ANALYSE_PSUB16x16;
220 if (avctx->partitions & X264_PART_P4X4)
221 x4->params.analyse.inter |= X264_ANALYSE_PSUB8x8;
222 if (avctx->partitions & X264_PART_B8X8)
223 x4->params.analyse.inter |= X264_ANALYSE_BSUB16x16;
224 }
225
226 x4->params.analyse.i_direct_mv_pred = avctx->directpred;
227
228 x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED;
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 if (avctx->weighted_p_pred >= 0)
322 x4->params.analyse.i_weighted_pred = avctx->weighted_p_pred;
323 #endif
324
325 if (x4->aq_mode >= 0)
326 x4->params.rc.i_aq_mode = x4->aq_mode;
327 if (x4->aq_strength >= 0)
328 x4->params.rc.f_aq_strength = x4->aq_strength;
329 if (x4->psy_rd >= 0)
330 x4->params.analyse.f_psy_rd = x4->psy_rd;
331 if (x4->psy_trellis >= 0)
332 x4->params.analyse.f_psy_trellis = x4->psy_trellis;
333 if (x4->rc_lookahead >= 0)
334 x4->params.rc.i_lookahead = x4->rc_lookahead;
335 if (x4->weightp >= 0)
336 x4->params.analyse.i_weighted_pred = x4->weightp;
337
338
339
340 if (x4->fastfirstpass)
341 x264_param_apply_fastfirstpass(&x4->params);
342
343 if (x4->profile)
344 if (x264_param_apply_profile(&x4->params, x4->profile) < 0) {
345 av_log(avctx, AV_LOG_ERROR, "Error setting profile %s.\n", x4->profile);
346 return AVERROR(EINVAL);
347 }
348
349 x4->params.i_width = avctx->width;
350 x4->params.i_height = avctx->height;
351 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
352 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
353 x4->params.i_fps_num = x4->params.i_timebase_den = avctx->time_base.den;
354 x4->params.i_fps_den = x4->params.i_timebase_num = avctx->time_base.num;
355
356 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
357 x4->params.analyse.b_ssim = avctx->flags2 & CODEC_FLAG2_SSIM;
358
359 x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD;
360
361 x4->params.i_threads = avctx->thread_count;
362
363 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
364
365 x4->params.b_open_gop = !(avctx->flags & CODEC_FLAG_CLOSED_GOP);
366
367 x4->params.i_slice_count = avctx->slices;
368
369 x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
370
371 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
372 x4->params.b_repeat_headers = 0;
373
374 // update AVCodecContext with x264 parameters
375 avctx->has_b_frames = x4->params.i_bframe ?
376 x4->params.i_bframe_pyramid ? 2 : 1 : 0;
377 avctx->bit_rate = x4->params.rc.i_bitrate*1000;
378 #if FF_API_X264_GLOBAL_OPTS
379 avctx->crf = x4->params.rc.f_rf_constant;
380 #endif
381
382 x4->enc = x264_encoder_open(&x4->params);
383 if (!x4->enc)
384 return -1;
385
386 avctx->coded_frame = &x4->out_pic;
387
388 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
389 x264_nal_t *nal;
390 int nnal, s, i;
391
392 s = x264_encoder_headers(x4->enc, &nal, &nnal);
393
394 for (i = 0; i < nnal; i++)
395 if (nal[i].i_type == NAL_SEI)
396 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
397
398 avctx->extradata = av_malloc(s);
399 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
400 }
401
402 return 0;
403 }
404
405 #define OFFSET(x) offsetof(X264Context, x)
406 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
407 static const AVOption options[] = {
408 { "preset", "Set the encoding preset (cf. x264 --fullhelp)", OFFSET(preset), FF_OPT_TYPE_STRING, { .str = "medium" }, 0, 0, VE},
409 { "tune", "Tune the encoding params (cf. x264 --fullhelp)", OFFSET(tune), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
410 { "profile", "Set profile restrictions (cf. x264 --fullhelp) ", OFFSET(profile), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
411 { "fastfirstpass", "Use fast settings when encoding first pass", OFFSET(fastfirstpass), FF_OPT_TYPE_INT, { 1 }, 0, 1, VE},
412 { "crf", "Select the quality for constant quality mode", OFFSET(crf), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
413 { "cqp", "Constant quantization parameter rate control method",OFFSET(cqp), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
414 { "aq_mode", "AQ method", OFFSET(aq_mode), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "aq_mode"},
415 { "none", NULL, 0, FF_OPT_TYPE_CONST, {X264_AQ_NONE}, INT_MIN, INT_MAX, VE, "aq_mode" },
416 { "variance", "Variance AQ (complexity mask)", 0, FF_OPT_TYPE_CONST, {X264_AQ_VARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
417 { "autovariance", "Auto-variance AQ (experimental)", 0, FF_OPT_TYPE_CONST, {X264_AQ_AUTOVARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
418 { "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},
419 { "pdy_rd", "Psy RD strength.", OFFSET(psy_rd), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
420 { "psy_trellis", "Psy trellis strength", OFFSET(psy_trellis), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
421 { "rc_lookahead", "Number of frames to look ahead for frametype and ratecontrol", OFFSET(rc_lookahead), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
422 { "weightp", "Weighted prediction analysis method.", OFFSET(weightp), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "weightp" },
423 { "none", NULL, 0, FF_OPT_TYPE_CONST, {X264_WEIGHTP_NONE}, INT_MIN, INT_MAX, VE, "weightp" },
424 { "simple", NULL, 0, FF_OPT_TYPE_CONST, {X264_WEIGHTP_SIMPLE}, INT_MIN, INT_MAX, VE, "weightp" },
425 { "smart", NULL, 0, FF_OPT_TYPE_CONST, {X264_WEIGHTP_SMART}, INT_MIN, INT_MAX, VE, "weightp" },
426 { NULL },
427 };
428
429 static const AVClass class = {
430 .class_name = "libx264",
431 .item_name = av_default_item_name,
432 .option = options,
433 .version = LIBAVUTIL_VERSION_INT,
434 };
435
436 static const AVCodecDefault x264_defaults[] = {
437 { "b", "0" },
438 { NULL },
439 };
440
441 AVCodec ff_libx264_encoder = {
442 .name = "libx264",
443 .type = AVMEDIA_TYPE_VIDEO,
444 .id = CODEC_ID_H264,
445 .priv_data_size = sizeof(X264Context),
446 .init = X264_init,
447 .encode = X264_frame,
448 .close = X264_close,
449 .capabilities = CODEC_CAP_DELAY,
450 .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_YUVJ420P, PIX_FMT_NONE },
451 .long_name = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
452 .priv_class = &class,
453 .defaults = x264_defaults,
454 };