libtheoraenc.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2006 Paul Richards <paul.richards@gmail.com>
3  *
4  * This file is part of Libav.
5  *
6  * Libav is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * Libav is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with Libav; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
33 /* Libav includes */
34 #include "libavutil/intreadwrite.h"
35 #include "libavutil/log.h"
36 #include "libavutil/base64.h"
37 #include "avcodec.h"
38 
39 /* libtheora includes */
40 #include <theora/theoraenc.h>
41 
42 typedef struct TheoraContext {
43  th_enc_ctx *t_state;
44  uint8_t *stats;
47  int uv_hshift;
48  int uv_vshift;
51 
53 static int concatenate_packet(unsigned int* offset,
54  AVCodecContext* avc_context,
55  const ogg_packet* packet)
56 {
57  const char* message = NULL;
58  uint8_t* newdata = NULL;
59  int newsize = avc_context->extradata_size + 2 + packet->bytes;
60 
61  if (packet->bytes < 0) {
62  message = "ogg_packet has negative size";
63  } else if (packet->bytes > 0xffff) {
64  message = "ogg_packet is larger than 65535 bytes";
65  } else if (newsize < avc_context->extradata_size) {
66  message = "extradata_size would overflow";
67  } else {
68  newdata = av_realloc(avc_context->extradata, newsize);
69  if (!newdata)
70  message = "av_realloc failed";
71  }
72  if (message) {
73  av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message);
74  return -1;
75  }
76 
77  avc_context->extradata = newdata;
78  avc_context->extradata_size = newsize;
79  AV_WB16(avc_context->extradata + (*offset), packet->bytes);
80  *offset += 2;
81  memcpy(avc_context->extradata + (*offset), packet->packet, packet->bytes);
82  (*offset) += packet->bytes;
83  return 0;
84 }
85 
86 static int get_stats(AVCodecContext *avctx, int eos)
87 {
88 #ifdef TH_ENCCTL_2PASS_OUT
89  TheoraContext *h = avctx->priv_data;
90  uint8_t *buf;
91  int bytes;
92 
93  bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_OUT, &buf, sizeof(buf));
94  if (bytes < 0) {
95  av_log(avctx, AV_LOG_ERROR, "Error getting first pass stats\n");
96  return -1;
97  }
98  if (!eos) {
100  h->stats_offset + bytes);
101  memcpy(h->stats + h->stats_offset, buf, bytes);
102  h->stats_offset += bytes;
103  } else {
104  int b64_size = AV_BASE64_SIZE(h->stats_offset);
105  // libtheora generates a summary header at the end
106  memcpy(h->stats, buf, bytes);
107  avctx->stats_out = av_malloc(b64_size);
108  av_base64_encode(avctx->stats_out, b64_size, h->stats, h->stats_offset);
109  }
110  return 0;
111 #else
112  av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n");
113  return -1;
114 #endif
115 }
116 
117 // libtheora won't read the entire buffer we give it at once, so we have to
118 // repeatedly submit it...
119 static int submit_stats(AVCodecContext *avctx)
120 {
121 #ifdef TH_ENCCTL_2PASS_IN
122  TheoraContext *h = avctx->priv_data;
123  int bytes;
124  if (!h->stats) {
125  if (!avctx->stats_in) {
126  av_log(avctx, AV_LOG_ERROR, "No statsfile for second pass\n");
127  return -1;
128  }
129  h->stats_size = strlen(avctx->stats_in) * 3/4;
130  h->stats = av_malloc(h->stats_size);
131  h->stats_size = av_base64_decode(h->stats, avctx->stats_in, h->stats_size);
132  }
133  while (h->stats_size - h->stats_offset > 0) {
134  bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_IN,
135  h->stats + h->stats_offset,
136  h->stats_size - h->stats_offset);
137  if (bytes < 0) {
138  av_log(avctx, AV_LOG_ERROR, "Error submitting stats\n");
139  return -1;
140  }
141  if (!bytes)
142  return 0;
143  h->stats_offset += bytes;
144  }
145  return 0;
146 #else
147  av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n");
148  return -1;
149 #endif
150 }
151 
152 static av_cold int encode_init(AVCodecContext* avc_context)
153 {
154  th_info t_info;
155  th_comment t_comment;
156  ogg_packet o_packet;
157  unsigned int offset;
158  TheoraContext *h = avc_context->priv_data;
159  uint32_t gop_size = avc_context->gop_size;
160 
161  /* Set up the theora_info struct */
162  th_info_init(&t_info);
163  t_info.frame_width = FFALIGN(avc_context->width, 16);
164  t_info.frame_height = FFALIGN(avc_context->height, 16);
165  t_info.pic_width = avc_context->width;
166  t_info.pic_height = avc_context->height;
167  t_info.pic_x = 0;
168  t_info.pic_y = 0;
169  /* Swap numerator and denominator as time_base in AVCodecContext gives the
170  * time period between frames, but theora_info needs the framerate. */
171  t_info.fps_numerator = avc_context->time_base.den;
172  t_info.fps_denominator = avc_context->time_base.num;
173  if (avc_context->sample_aspect_ratio.num) {
174  t_info.aspect_numerator = avc_context->sample_aspect_ratio.num;
175  t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;
176  } else {
177  t_info.aspect_numerator = 1;
178  t_info.aspect_denominator = 1;
179  }
180 
181  if (avc_context->color_primaries == AVCOL_PRI_BT470M)
182  t_info.colorspace = TH_CS_ITU_REC_470M;
183  else if (avc_context->color_primaries == AVCOL_PRI_BT470BG)
184  t_info.colorspace = TH_CS_ITU_REC_470BG;
185  else
186  t_info.colorspace = TH_CS_UNSPECIFIED;
187 
188  if (avc_context->pix_fmt == PIX_FMT_YUV420P)
189  t_info.pixel_fmt = TH_PF_420;
190  else if (avc_context->pix_fmt == PIX_FMT_YUV422P)
191  t_info.pixel_fmt = TH_PF_422;
192  else if (avc_context->pix_fmt == PIX_FMT_YUV444P)
193  t_info.pixel_fmt = TH_PF_444;
194  else {
195  av_log(avc_context, AV_LOG_ERROR, "Unsupported pix_fmt\n");
196  return -1;
197  }
199 
200  if (avc_context->flags & CODEC_FLAG_QSCALE) {
201  /* to be constant with the libvorbis implementation, clip global_quality to 0 - 10
202  Theora accepts a quality parameter p, which is:
203  * 0 <= p <=63
204  * an int value
205  */
206  t_info.quality = av_clip(avc_context->global_quality / (float)FF_QP2LAMBDA, 0, 10) * 6.3;
207  t_info.target_bitrate = 0;
208  } else {
209  t_info.target_bitrate = avc_context->bit_rate;
210  t_info.quality = 0;
211  }
212 
213  /* Now initialise libtheora */
214  h->t_state = th_encode_alloc(&t_info);
215  if (!h->t_state) {
216  av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n");
217  return -1;
218  }
219 
220  h->keyframe_mask = (1 << t_info.keyframe_granule_shift) - 1;
221  /* Clear up theora_info struct */
222  th_info_clear(&t_info);
223 
224  if (th_encode_ctl(h->t_state, TH_ENCCTL_SET_KEYFRAME_FREQUENCY_FORCE,
225  &gop_size, sizeof(gop_size))) {
226  av_log(avc_context, AV_LOG_ERROR, "Error setting GOP size\n");
227  return -1;
228  }
229 
230  // need to enable 2 pass (via TH_ENCCTL_2PASS_) before encoding headers
231  if (avc_context->flags & CODEC_FLAG_PASS1) {
232  if (get_stats(avc_context, 0))
233  return -1;
234  } else if (avc_context->flags & CODEC_FLAG_PASS2) {
235  if (submit_stats(avc_context))
236  return -1;
237  }
238 
239  /*
240  Output first header packet consisting of theora
241  header, comment, and tables.
242 
243  Each one is prefixed with a 16bit size, then they
244  are concatenated together into libavcodec's extradata.
245  */
246  offset = 0;
247 
248  /* Headers */
249  th_comment_init(&t_comment);
250 
251  while (th_encode_flushheader(h->t_state, &t_comment, &o_packet))
252  if (concatenate_packet(&offset, avc_context, &o_packet))
253  return -1;
254 
255  th_comment_clear(&t_comment);
256 
257  /* Set up the output AVFrame */
258  avc_context->coded_frame= avcodec_alloc_frame();
259 
260  return 0;
261 }
262 
263 static int encode_frame(AVCodecContext* avc_context, uint8_t *outbuf,
264  int buf_size, void *data)
265 {
266  th_ycbcr_buffer t_yuv_buffer;
267  TheoraContext *h = avc_context->priv_data;
268  AVFrame *frame = data;
269  ogg_packet o_packet;
270  int result, i;
271 
272  // EOS, finish and get 1st pass stats if applicable
273  if (!frame) {
274  th_encode_packetout(h->t_state, 1, &o_packet);
275  if (avc_context->flags & CODEC_FLAG_PASS1)
276  if (get_stats(avc_context, 1))
277  return -1;
278  return 0;
279  }
280 
281  /* Copy planes to the theora yuv_buffer */
282  for (i = 0; i < 3; i++) {
283  t_yuv_buffer[i].width = FFALIGN(avc_context->width, 16) >> (i && h->uv_hshift);
284  t_yuv_buffer[i].height = FFALIGN(avc_context->height, 16) >> (i && h->uv_vshift);
285  t_yuv_buffer[i].stride = frame->linesize[i];
286  t_yuv_buffer[i].data = frame->data[i];
287  }
288 
289  if (avc_context->flags & CODEC_FLAG_PASS2)
290  if (submit_stats(avc_context))
291  return -1;
292 
293  /* Now call into theora_encode_YUVin */
294  result = th_encode_ycbcr_in(h->t_state, t_yuv_buffer);
295  if (result) {
296  const char* message;
297  switch (result) {
298  case -1:
299  message = "differing frame sizes";
300  break;
301  case TH_EINVAL:
302  message = "encoder is not ready or is finished";
303  break;
304  default:
305  message = "unknown reason";
306  break;
307  }
308  av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result);
309  return -1;
310  }
311 
312  if (avc_context->flags & CODEC_FLAG_PASS1)
313  if (get_stats(avc_context, 0))
314  return -1;
315 
316  /* Pick up returned ogg_packet */
317  result = th_encode_packetout(h->t_state, 0, &o_packet);
318  switch (result) {
319  case 0:
320  /* No packet is ready */
321  return 0;
322  case 1:
323  /* Success, we have a packet */
324  break;
325  default:
326  av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result);
327  return -1;
328  }
329 
330  /* Copy ogg_packet content out to buffer */
331  if (buf_size < o_packet.bytes) {
332  av_log(avc_context, AV_LOG_ERROR, "encoded frame too large\n");
333  return -1;
334  }
335  memcpy(outbuf, o_packet.packet, o_packet.bytes);
336 
337  // HACK: assumes no encoder delay, this is true until libtheora becomes
338  // multithreaded (which will be disabled unless explictly requested)
339  avc_context->coded_frame->pts = frame->pts;
340  avc_context->coded_frame->key_frame = !(o_packet.granulepos & h->keyframe_mask);
341 
342  return o_packet.bytes;
343 }
344 
345 static av_cold int encode_close(AVCodecContext* avc_context)
346 {
347  TheoraContext *h = avc_context->priv_data;
348 
349  th_encode_free(h->t_state);
350  av_freep(&h->stats);
351  av_freep(&avc_context->coded_frame);
352  av_freep(&avc_context->stats_out);
353  av_freep(&avc_context->extradata);
354  avc_context->extradata_size = 0;
355 
356  return 0;
357 }
358 
361  .name = "libtheora",
362  .type = AVMEDIA_TYPE_VIDEO,
363  .id = CODEC_ID_THEORA,
364  .priv_data_size = sizeof(TheoraContext),
365  .init = encode_init,
366  .close = encode_close,
367  .encode = encode_frame,
368  .capabilities = CODEC_CAP_DELAY, // needed to get the statsfile summary
370  .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"),
371 };