FFmpeg: libavcodec/libtheoraenc.c Source File

FFmpeg

[フレーム]

libavcodec

libtheoraenc.c

Go to the documentation of this file.

1 /*

3 *

4 * This file is part of FFmpeg.

5 *

6 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software

18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

19 */

21 /**

22 * @file

23 * @brief Theora encoder using libtheora.

24 * @author Paul Richards <paul.richards@gmail.com>

25 *

26 * A lot of this is copy / paste from other output codecs in

27 * libavcodec or pure guesswork (or both).

28 *

29 * I have used t_ prefixes on variables which are libtheora types

30 * and o_ prefixes on variables which are libogg types.

31 */

33 /* FFmpeg includes */

34 #include "libavutil/common.h"

35 #include "libavutil/intreadwrite.h"

36 #include "libavutil/pixdesc.h"

37 #include "libavutil/log.h"

38 #include "libavutil/base64.h"

39 #include "avcodec.h"

40 #include "internal.h"

42 /* libtheora includes */

43 #include <theora/theoraenc.h>

45 typedef struct TheoraContext {

46 th_enc_ctx *t_state;

47 uint8_t *stats;

48 int stats_size;

49 int stats_offset;

50 int uv_hshift;

51 int uv_vshift;

52 int keyframe_mask;

53 } TheoraContext;

55 /** Concatenate an ogg_packet into the extradata. */

56 static int concatenate_packet(unsigned int* offset,

57 AVCodecContext* avc_context,

58 const ogg_packet* packet)

59 {

60 const char* message = NULL;

61 int newsize = avc_context->extradata_size + 2 + packet->bytes;

62 int err = AVERROR_INVALIDDATA;

64 if (packet->bytes < 0) {

65 message = "ogg_packet has negative size";

66 } else if (packet->bytes > 0xffff) {

67 message = "ogg_packet is larger than 65535 bytes";

68 } else if (newsize < avc_context->extradata_size) {

69 message = "extradata_size would overflow";

70 } else {

71 if ((err = av_reallocp(&avc_context->extradata, newsize)) < 0) {

72 avc_context->extradata_size = 0;

73 message = "av_realloc failed";

74 }

75 }

76 if (message) {

77 av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message);

78 return err;

79 }

81 avc_context->extradata_size = newsize;

82 AV_WB16(avc_context->extradata + (*offset), packet->bytes);

83 *offset += 2;

84 memcpy(avc_context->extradata + (*offset), packet->packet, packet->bytes);

85 (*offset) += packet->bytes;

86 return 0;

87 }

89 static int get_stats(AVCodecContext *avctx, int eos)

90 {

91 #ifdef TH_ENCCTL_2PASS_OUT

92 TheoraContext *h = avctx->priv_data;

93 uint8_t *buf;

94 int bytes;

96 bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_OUT, &buf, sizeof(buf));

97 if (bytes < 0) {

98 av_log(avctx, AV_LOG_ERROR, "Error getting first pass stats\n");

99 return AVERROR_EXTERNAL;

100 }

101 if (!eos) {

102 void *tmp = av_fast_realloc(h->stats, &h->stats_size,

103 h->stats_offset + bytes);

104 if (!tmp)

105 return AVERROR(ENOMEM);

106 h->stats = tmp;

107 memcpy(h->stats + h->stats_offset, buf, bytes);

108 h->stats_offset += bytes;

109 } else {

110 int b64_size = AV_BASE64_SIZE(h->stats_offset);

111 // libtheora generates a summary header at the end

112 memcpy(h->stats, buf, bytes);

113 avctx->stats_out = av_malloc(b64_size);

114 av_base64_encode(avctx->stats_out, b64_size, h->stats, h->stats_offset);

115 }

116 return 0;

117 #else

118 av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n");

119 return AVERROR(ENOSUP);

120 #endif

121 }

122

123 // libtheora won't read the entire buffer we give it at once, so we have to

124 // repeatedly submit it...

125 static int submit_stats(AVCodecContext *avctx)

126 {

127 #ifdef TH_ENCCTL_2PASS_IN

128 TheoraContext *h = avctx->priv_data;

129 int bytes;

130 if (!h->stats) {

131 if (!avctx->stats_in) {

132 av_log(avctx, AV_LOG_ERROR, "No statsfile for second pass\n");

133 return AVERROR(EINVAL);

134 }

135 h->stats_size = strlen(avctx->stats_in) * 3/4;

136 h->stats = av_malloc(h->stats_size);

137 if (!h->stats) {

138 h->stats_size = 0;

139 return AVERROR(ENOMEM);

140 }

141 h->stats_size = av_base64_decode(h->stats, avctx->stats_in, h->stats_size);

142 }

143 while (h->stats_size - h->stats_offset > 0) {

144 bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_IN,

145 h->stats + h->stats_offset,

146 h->stats_size - h->stats_offset);

147 if (bytes < 0) {

148 av_log(avctx, AV_LOG_ERROR, "Error submitting stats\n");

149 return AVERROR_EXTERNAL;

150 }

151 if (!bytes)

152 return 0;

153 h->stats_offset += bytes;

154 }

155 return 0;

156 #else

157 av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n");

158 return AVERROR(ENOSUP);

159 #endif

160 }

161

162 static av_cold int encode_init(AVCodecContext* avc_context)

163 {

164 th_info t_info;

165 th_comment t_comment;

166 ogg_packet o_packet;

167 unsigned int offset;

168 TheoraContext *h = avc_context->priv_data;

169 uint32_t gop_size = avc_context->gop_size;

170 int ret;

171

172 /* Set up the theora_info struct */

173 th_info_init(&t_info);

174 t_info.frame_width = FFALIGN(avc_context->width, 16);

175 t_info.frame_height = FFALIGN(avc_context->height, 16);

176 t_info.pic_width = avc_context->width;

177 t_info.pic_height = avc_context->height;

178 t_info.pic_x = 0;

179 t_info.pic_y = 0;

180 /* Swap numerator and denominator as time_base in AVCodecContext gives the

181 * time period between frames, but theora_info needs the framerate. */

182 t_info.fps_numerator = avc_context->time_base.den;

183 t_info.fps_denominator = avc_context->time_base.num;

184 if (avc_context->sample_aspect_ratio.num) {

185 t_info.aspect_numerator = avc_context->sample_aspect_ratio.num;

186 t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;

187 } else {

188 t_info.aspect_numerator = 1;

189 t_info.aspect_denominator = 1;

190 }

191

192 if (avc_context->color_primaries == AVCOL_PRI_BT470M)

193 t_info.colorspace = TH_CS_ITU_REC_470M;

194 else if (avc_context->color_primaries == AVCOL_PRI_BT470BG)

195 t_info.colorspace = TH_CS_ITU_REC_470BG;

196 else

197 t_info.colorspace = TH_CS_UNSPECIFIED;

198

199 if (avc_context->pix_fmt == AV_PIX_FMT_YUV420P)

200 t_info.pixel_fmt = TH_PF_420;

201 else if (avc_context->pix_fmt == AV_PIX_FMT_YUV422P)

202 t_info.pixel_fmt = TH_PF_422;

203 else if (avc_context->pix_fmt == AV_PIX_FMT_YUV444P)

204 t_info.pixel_fmt = TH_PF_444;

205 else {

206 av_log(avc_context, AV_LOG_ERROR, "Unsupported pix_fmt\n");

207 return AVERROR(EINVAL);

208 }

209 avcodec_get_chroma_sub_sample(avc_context->pix_fmt, &h->uv_hshift, &h->uv_vshift);

210

211 if (avc_context->flags & CODEC_FLAG_QSCALE) {

212 /* Clip global_quality in QP units to the [0 - 10] range

213 to be consistent with the libvorbis implementation.

214 Theora accepts a quality parameter which is an int value in

215 the [0 - 63] range.

216 */

217 t_info.quality = av_clipf(avc_context->global_quality / (float)FF_QP2LAMBDA, 0, 10) * 6.3;

218 t_info.target_bitrate = 0;

219 } else {

220 t_info.target_bitrate = avc_context->bit_rate;

221 t_info.quality = 0;

222 }

223

224 /* Now initialise libtheora */

225 h->t_state = th_encode_alloc(&t_info);

226 if (!h->t_state) {

227 av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n");

228 return AVERROR_EXTERNAL;

229 }

230

231 h->keyframe_mask = (1 << t_info.keyframe_granule_shift) - 1;

232 /* Clear up theora_info struct */

233 th_info_clear(&t_info);

234

235 if (th_encode_ctl(h->t_state, TH_ENCCTL_SET_KEYFRAME_FREQUENCY_FORCE,

236 &gop_size, sizeof(gop_size))) {

237 av_log(avc_context, AV_LOG_ERROR, "Error setting GOP size\n");

238 return AVERROR_EXTERNAL;

239 }

240

241 // need to enable 2 pass (via TH_ENCCTL_2PASS_) before encoding headers

242 if (avc_context->flags & CODEC_FLAG_PASS1) {

243 if ((ret = get_stats(avc_context, 0)) < 0)

244 return ret;

245 } else if (avc_context->flags & CODEC_FLAG_PASS2) {

246 if ((ret = submit_stats(avc_context)) < 0)

247 return ret;

248 }

249

250 /*

251 Output first header packet consisting of theora

252 header, comment, and tables.

253

254 Each one is prefixed with a 16bit size, then they

255 are concatenated together into libavcodec's extradata.

256 */

257 offset = 0;

258

259 /* Headers */

260 th_comment_init(&t_comment);

261

262 while (th_encode_flushheader(h->t_state, &t_comment, &o_packet))

263 if ((ret = concatenate_packet(&offset, avc_context, &o_packet)) < 0)

264 return ret;

265

266 th_comment_clear(&t_comment);

267

268 /* Set up the output AVFrame */

269 avc_context->coded_frame = av_frame_alloc();

270

271 return 0;

272 }

273

274 static int encode_frame(AVCodecContext* avc_context, AVPacket *pkt,

275 const AVFrame *frame, int *got_packet)

276 {

277 th_ycbcr_buffer t_yuv_buffer;

278 TheoraContext *h = avc_context->priv_data;

279 ogg_packet o_packet;

280 int result, i, ret;

281

282 // EOS, finish and get 1st pass stats if applicable

283 if (!frame) {

284 th_encode_packetout(h->t_state, 1, &o_packet);

285 if (avc_context->flags & CODEC_FLAG_PASS1)

286 if ((ret = get_stats(avc_context, 1)) < 0)

287 return ret;

288 return 0;

289 }

290

291 /* Copy planes to the theora yuv_buffer */

292 for (i = 0; i < 3; i++) {

293 t_yuv_buffer[i].width = FFALIGN(avc_context->width, 16) >> (i && h->uv_hshift);

294 t_yuv_buffer[i].height = FFALIGN(avc_context->height, 16) >> (i && h->uv_vshift);

295 t_yuv_buffer[i].stride = frame->linesize[i];

296 t_yuv_buffer[i].data = frame->data[i];

297 }

298

299 if (avc_context->flags & CODEC_FLAG_PASS2)

300 if ((ret = submit_stats(avc_context)) < 0)

301 return ret;

302

303 /* Now call into theora_encode_YUVin */

304 result = th_encode_ycbcr_in(h->t_state, t_yuv_buffer);

305 if (result) {

306 const char* message;

307 switch (result) {

308 case -1:

309 message = "differing frame sizes";

310 break;

311 case TH_EINVAL:

312 message = "encoder is not ready or is finished";

313 break;

314 default:

315 message = "unknown reason";

316 break;

317 }

318 av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result);

319 return AVERROR_EXTERNAL;

320 }

321

322 if (avc_context->flags & CODEC_FLAG_PASS1)

323 if ((ret = get_stats(avc_context, 0)) < 0)

324 return ret;

325

326 /* Pick up returned ogg_packet */

327 result = th_encode_packetout(h->t_state, 0, &o_packet);

328 switch (result) {

329 case 0:

330 /* No packet is ready */

331 return 0;

332 case 1:

333 /* Success, we have a packet */

334 break;

335 default:

336 av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result);

337 return AVERROR_EXTERNAL;

338 }

339

340 /* Copy ogg_packet content out to buffer */

341 if ((ret = ff_alloc_packet2(avc_context, pkt, o_packet.bytes)) < 0)

342 return ret;

343 memcpy(pkt->data, o_packet.packet, o_packet.bytes);

344

345 // HACK: assumes no encoder delay, this is true until libtheora becomes

346 // multithreaded (which will be disabled unless explicitly requested)

347 pkt->pts = pkt->dts = frame->pts;

348 avc_context->coded_frame->key_frame = !(o_packet.granulepos & h->keyframe_mask);

349 if (avc_context->coded_frame->key_frame)

350 pkt->flags |= AV_PKT_FLAG_KEY;

351 *got_packet = 1;

352

353 return 0;

354 }

355

356 static av_cold int encode_close(AVCodecContext* avc_context)

357 {

358 TheoraContext *h = avc_context->priv_data;

359

360 th_encode_free(h->t_state);

361 av_freep(&h->stats);

362 av_frame_free(&avc_context->coded_frame);

363 av_freep(&avc_context->stats_out);

364 av_freep(&avc_context->extradata);

365 avc_context->extradata_size = 0;

366

367 return 0;

368 }

369

370 /** AVCodec struct exposed to libavcodec */

371 AVCodec ff_libtheora_encoder = {

372 .name = "libtheora",

373 .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"),

374 .type = AVMEDIA_TYPE_VIDEO,

375 .id = AV_CODEC_ID_THEORA,

376 .priv_data_size = sizeof(TheoraContext),

377 .init = encode_init,

378 .close = encode_close,

379 .encode2 = encode_frame,

380 .capabilities = CODEC_CAP_DELAY, // needed to get the statsfile summary

381 .pix_fmts = (const enum AVPixelFormat[]){

382 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_NONE

383 },

384 };

Generated on Sun Mar 8 2015 02:34:55 for FFmpeg by doxygen 1.8.2