Libav 0.7.1
|
00001 /* 00002 * General DV muxer/demuxer 00003 * Copyright (c) 2003 Roman Shaposhnik 00004 * 00005 * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth 00006 * of DV technical info. 00007 * 00008 * Raw DV format 00009 * Copyright (c) 2002 Fabrice Bellard 00010 * 00011 * 50 Mbps (DVCPRO50) and 100 Mbps (DVCPRO HD) support 00012 * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com> 00013 * Funded by BBC Research & Development 00014 * 00015 * This file is part of Libav. 00016 * 00017 * Libav is free software; you can redistribute it and/or 00018 * modify it under the terms of the GNU Lesser General Public 00019 * License as published by the Free Software Foundation; either 00020 * version 2.1 of the License, or (at your option) any later version. 00021 * 00022 * Libav is distributed in the hope that it will be useful, 00023 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00024 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00025 * Lesser General Public License for more details. 00026 * 00027 * You should have received a copy of the GNU Lesser General Public 00028 * License along with Libav; if not, write to the Free Software 00029 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00030 */ 00031 #include <time.h> 00032 #include "avformat.h" 00033 #include "libavcodec/dvdata.h" 00034 #include "libavutil/intreadwrite.h" 00035 #include "dv.h" 00036 00037 struct DVDemuxContext { 00038 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */ 00039 AVFormatContext* fctx; 00040 AVStream* vst; 00041 AVStream* ast[4]; 00042 AVPacket audio_pkt[4]; 00043 uint8_t audio_buf[4][8192]; 00044 int ach; 00045 int frames; 00046 uint64_t abytes; 00047 }; 00048 00049 static inline uint16_t dv_audio_12to16(uint16_t sample) 00050 { 00051 uint16_t shift, result; 00052 00053 sample = (sample < 0x800) ? sample : sample | 0xf000; 00054 shift = (sample & 0xf00) >> 8; 00055 00056 if (shift < 0x2 || shift > 0xd) { 00057 result = sample; 00058 } else if (shift < 0x8) { 00059 shift--; 00060 result = (sample - (256 * shift)) << shift; 00061 } else { 00062 shift = 0xe - shift; 00063 result = ((sample + ((256 * shift) + 1)) << shift) - 1; 00064 } 00065 00066 return result; 00067 } 00068 00069 /* 00070 * This is the dumbest implementation of all -- it simply looks at 00071 * a fixed offset and if pack isn't there -- fails. We might want 00072 * to have a fallback mechanism for complete search of missing packs. 00073 */ 00074 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t) 00075 { 00076 int offs; 00077 00078 switch (t) { 00079 case dv_audio_source: 00080 offs = (80*6 + 80*16*3 + 3); 00081 break; 00082 case dv_audio_control: 00083 offs = (80*6 + 80*16*4 + 3); 00084 break; 00085 case dv_video_control: 00086 offs = (80*5 + 48 + 5); 00087 break; 00088 default: 00089 return NULL; 00090 } 00091 00092 return frame[offs] == t ? &frame[offs] : NULL; 00093 } 00094 00095 /* 00096 * There's a couple of assumptions being made here: 00097 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples. 00098 * We can pass them upwards when ffmpeg will be ready to deal with them. 00099 * 2. We don't do software emphasis. 00100 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples 00101 * are converted into 16bit linear ones. 00102 */ 00103 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4], 00104 const DVprofile *sys) 00105 { 00106 int size, chan, i, j, d, of, smpls, freq, quant, half_ch; 00107 uint16_t lc, rc; 00108 const uint8_t* as_pack; 00109 uint8_t *pcm, ipcm; 00110 00111 as_pack = dv_extract_pack(frame, dv_audio_source); 00112 if (!as_pack) /* No audio ? */ 00113 return 0; 00114 00115 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */ 00116 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */ 00117 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */ 00118 00119 if (quant > 1) 00120 return -1; /* unsupported quantization */ 00121 00122 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) 00123 return AVERROR_INVALIDDATA; 00124 00125 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */ 00126 half_ch = sys->difseg_size / 2; 00127 00128 /* We work with 720p frames split in half, thus even frames have 00129 * channels 0,1 and odd 2,3. */ 00130 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0; 00131 00132 /* for each DIF channel */ 00133 for (chan = 0; chan < sys->n_difchan; chan++) { 00134 /* next stereo channel (50Mbps and 100Mbps only) */ 00135 pcm = ppcm[ipcm++]; 00136 if (!pcm) 00137 break; 00138 00139 /* for each DIF segment */ 00140 for (i = 0; i < sys->difseg_size; i++) { 00141 frame += 6 * 80; /* skip DIF segment header */ 00142 if (quant == 1 && i == half_ch) { 00143 /* next stereo channel (12bit mode only) */ 00144 pcm = ppcm[ipcm++]; 00145 if (!pcm) 00146 break; 00147 } 00148 00149 /* for each AV sequence */ 00150 for (j = 0; j < 9; j++) { 00151 for (d = 8; d < 80; d += 2) { 00152 if (quant == 0) { /* 16bit quantization */ 00153 of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride; 00154 if (of*2 >= size) 00155 continue; 00156 00157 pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit 00158 pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM 00159 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00) 00160 pcm[of*2+1] = 0; 00161 } else { /* 12bit quantization */ 00162 lc = ((uint16_t)frame[d] << 4) | 00163 ((uint16_t)frame[d+2] >> 4); 00164 rc = ((uint16_t)frame[d+1] << 4) | 00165 ((uint16_t)frame[d+2] & 0x0f); 00166 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc)); 00167 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc)); 00168 00169 of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride; 00170 if (of*2 >= size) 00171 continue; 00172 00173 pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit 00174 pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM 00175 of = sys->audio_shuffle[i%half_ch+half_ch][j] + 00176 (d - 8) / 3 * sys->audio_stride; 00177 pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit 00178 pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM 00179 ++d; 00180 } 00181 } 00182 00183 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */ 00184 } 00185 } 00186 } 00187 00188 return size; 00189 } 00190 00191 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame) 00192 { 00193 const uint8_t* as_pack; 00194 int freq, stype, smpls, quant, i, ach; 00195 00196 as_pack = dv_extract_pack(frame, dv_audio_source); 00197 if (!as_pack || !c->sys) { /* No audio ? */ 00198 c->ach = 0; 00199 return 0; 00200 } 00201 00202 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */ 00203 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */ 00204 stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */ 00205 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */ 00206 00207 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) { 00208 av_log(c->fctx, AV_LOG_ERROR, 00209 "Unrecognized audio sample rate index (%d)\n", freq); 00210 return 0; 00211 } 00212 00213 if (stype > 3) { 00214 av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype); 00215 c->ach = 0; 00216 return 0; 00217 } 00218 00219 /* note: ach counts PAIRS of channels (i.e. stereo channels) */ 00220 ach = ((int[4]){ 1, 0, 2, 4})[stype]; 00221 if (ach == 1 && quant && freq == 2) 00222 ach = 2; 00223 00224 /* Dynamic handling of the audio streams in DV */ 00225 for (i = 0; i < ach; i++) { 00226 if (!c->ast[i]) { 00227 c->ast[i] = av_new_stream(c->fctx, 0); 00228 if (!c->ast[i]) 00229 break; 00230 av_set_pts_info(c->ast[i], 64, 1, 30000); 00231 c->ast[i]->codec->codec_type = AVMEDIA_TYPE_AUDIO; 00232 c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE; 00233 00234 av_init_packet(&c->audio_pkt[i]); 00235 c->audio_pkt[i].size = 0; 00236 c->audio_pkt[i].data = c->audio_buf[i]; 00237 c->audio_pkt[i].stream_index = c->ast[i]->index; 00238 c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY; 00239 } 00240 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq]; 00241 c->ast[i]->codec->channels = 2; 00242 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16; 00243 c->ast[i]->start_time = 0; 00244 } 00245 c->ach = i; 00246 00247 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */; 00248 } 00249 00250 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame) 00251 { 00252 const uint8_t* vsc_pack; 00253 AVCodecContext* avctx; 00254 int apt, is16_9; 00255 int size = 0; 00256 00257 if (c->sys) { 00258 avctx = c->vst->codec; 00259 00260 av_set_pts_info(c->vst, 64, c->sys->time_base.num, 00261 c->sys->time_base.den); 00262 avctx->time_base= c->sys->time_base; 00263 if (!avctx->width){ 00264 avctx->width = c->sys->width; 00265 avctx->height = c->sys->height; 00266 } 00267 avctx->pix_fmt = c->sys->pix_fmt; 00268 00269 /* finding out SAR is a little bit messy */ 00270 vsc_pack = dv_extract_pack(frame, dv_video_control); 00271 apt = frame[4] & 0x07; 00272 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 || 00273 (!apt && (vsc_pack[2] & 0x07) == 0x07))); 00274 c->vst->sample_aspect_ratio = c->sys->sar[is16_9]; 00275 avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1}, 00276 c->sys->time_base); 00277 size = c->sys->frame_size; 00278 } 00279 return size; 00280 } 00281 00282 /* 00283 * The following 3 functions constitute our interface to the world 00284 */ 00285 00286 DVDemuxContext* dv_init_demux(AVFormatContext *s) 00287 { 00288 DVDemuxContext *c; 00289 00290 c = av_mallocz(sizeof(DVDemuxContext)); 00291 if (!c) 00292 return NULL; 00293 00294 c->vst = av_new_stream(s, 0); 00295 if (!c->vst) { 00296 av_free(c); 00297 return NULL; 00298 } 00299 00300 c->sys = NULL; 00301 c->fctx = s; 00302 memset(c->ast, 0, sizeof(c->ast)); 00303 c->ach = 0; 00304 c->frames = 0; 00305 c->abytes = 0; 00306 00307 c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO; 00308 c->vst->codec->codec_id = CODEC_ID_DVVIDEO; 00309 c->vst->codec->bit_rate = 25000000; 00310 c->vst->start_time = 0; 00311 00312 return c; 00313 } 00314 00315 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt) 00316 { 00317 int size = -1; 00318 int i; 00319 00320 for (i = 0; i < c->ach; i++) { 00321 if (c->ast[i] && c->audio_pkt[i].size) { 00322 *pkt = c->audio_pkt[i]; 00323 c->audio_pkt[i].size = 0; 00324 size = pkt->size; 00325 break; 00326 } 00327 } 00328 00329 return size; 00330 } 00331 00332 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt, 00333 uint8_t* buf, int buf_size) 00334 { 00335 int size, i; 00336 uint8_t *ppcm[4] = {0}; 00337 00338 if (buf_size < DV_PROFILE_BYTES || 00339 !(c->sys = ff_dv_frame_profile(c->sys, buf, buf_size)) || 00340 buf_size < c->sys->frame_size) { 00341 return -1; /* Broken frame, or not enough data */ 00342 } 00343 00344 /* Queueing audio packet */ 00345 /* FIXME: in case of no audio/bad audio we have to do something */ 00346 size = dv_extract_audio_info(c, buf); 00347 for (i = 0; i < c->ach; i++) { 00348 c->audio_pkt[i].size = size; 00349 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate; 00350 ppcm[i] = c->audio_buf[i]; 00351 } 00352 if (c->ach) 00353 dv_extract_audio(buf, ppcm, c->sys); 00354 00355 /* We work with 720p frames split in half, thus even frames have 00356 * channels 0,1 and odd 2,3. */ 00357 if (c->sys->height == 720) { 00358 if (buf[1] & 0x0C) { 00359 c->audio_pkt[2].size = c->audio_pkt[3].size = 0; 00360 } else { 00361 c->audio_pkt[0].size = c->audio_pkt[1].size = 0; 00362 c->abytes += size; 00363 } 00364 } else { 00365 c->abytes += size; 00366 } 00367 00368 /* Now it's time to return video packet */ 00369 size = dv_extract_video_info(c, buf); 00370 av_init_packet(pkt); 00371 pkt->data = buf; 00372 pkt->size = size; 00373 pkt->flags |= AV_PKT_FLAG_KEY; 00374 pkt->stream_index = c->vst->id; 00375 pkt->pts = c->frames; 00376 00377 c->frames++; 00378 00379 return size; 00380 } 00381 00382 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c, 00383 int64_t timestamp, int flags) 00384 { 00385 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk) 00386 const DVprofile* sys = ff_dv_codec_profile(c->vst->codec); 00387 int64_t offset; 00388 int64_t size = avio_size(s->pb) - s->data_offset; 00389 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size; 00390 00391 offset = sys->frame_size * timestamp; 00392 00393 if (size >= 0 && offset > max_offset) offset = max_offset; 00394 else if (offset < 0) offset = 0; 00395 00396 return offset + s->data_offset; 00397 } 00398 00399 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset) 00400 { 00401 c->frames= frame_offset; 00402 if (c->ach) 00403 c->abytes= av_rescale_q(c->frames, c->sys->time_base, 00404 (AVRational){8, c->ast[0]->codec->bit_rate}); 00405 c->audio_pkt[0].size = c->audio_pkt[1].size = 0; 00406 c->audio_pkt[2].size = c->audio_pkt[3].size = 0; 00407 } 00408 00409 /************************************************************ 00410 * Implementation of the easiest DV storage of all -- raw DV. 00411 ************************************************************/ 00412 00413 typedef struct RawDVContext { 00414 DVDemuxContext* dv_demux; 00415 uint8_t buf[DV_MAX_FRAME_SIZE]; 00416 } RawDVContext; 00417 00418 static int dv_read_header(AVFormatContext *s, 00419 AVFormatParameters *ap) 00420 { 00421 unsigned state, marker_pos = 0; 00422 RawDVContext *c = s->priv_data; 00423 00424 c->dv_demux = dv_init_demux(s); 00425 if (!c->dv_demux) 00426 return -1; 00427 00428 state = avio_rb32(s->pb); 00429 while ((state & 0xffffff7f) != 0x1f07003f) { 00430 if (s->pb->eof_reached) { 00431 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n"); 00432 return -1; 00433 } 00434 if (state == 0x003f0700 || state == 0xff3f0700) 00435 marker_pos = avio_tell(s->pb); 00436 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) { 00437 avio_seek(s->pb, -163, SEEK_CUR); 00438 state = avio_rb32(s->pb); 00439 break; 00440 } 00441 state = (state << 8) | avio_r8(s->pb); 00442 } 00443 AV_WB32(c->buf, state); 00444 00445 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 || 00446 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0) 00447 return AVERROR(EIO); 00448 00449 c->dv_demux->sys = ff_dv_frame_profile(c->dv_demux->sys, c->buf, DV_PROFILE_BYTES); 00450 if (!c->dv_demux->sys) { 00451 av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n"); 00452 return -1; 00453 } 00454 00455 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1}, 00456 c->dv_demux->sys->time_base); 00457 00458 return 0; 00459 } 00460 00461 00462 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt) 00463 { 00464 int size; 00465 RawDVContext *c = s->priv_data; 00466 00467 size = dv_get_packet(c->dv_demux, pkt); 00468 00469 if (size < 0) { 00470 if (!c->dv_demux->sys) 00471 return AVERROR(EIO); 00472 size = c->dv_demux->sys->frame_size; 00473 if (avio_read(s->pb, c->buf, size) <= 0) 00474 return AVERROR(EIO); 00475 00476 size = dv_produce_packet(c->dv_demux, pkt, c->buf, size); 00477 } 00478 00479 return size; 00480 } 00481 00482 static int dv_read_seek(AVFormatContext *s, int stream_index, 00483 int64_t timestamp, int flags) 00484 { 00485 RawDVContext *r = s->priv_data; 00486 DVDemuxContext *c = r->dv_demux; 00487 int64_t offset = dv_frame_offset(s, c, timestamp, flags); 00488 00489 dv_offset_reset(c, offset / c->sys->frame_size); 00490 00491 offset = avio_seek(s->pb, offset, SEEK_SET); 00492 return (offset < 0) ? offset : 0; 00493 } 00494 00495 static int dv_read_close(AVFormatContext *s) 00496 { 00497 RawDVContext *c = s->priv_data; 00498 av_free(c->dv_demux); 00499 return 0; 00500 } 00501 00502 static int dv_probe(AVProbeData *p) 00503 { 00504 unsigned state, marker_pos = 0; 00505 int i; 00506 int matches = 0; 00507 int secondary_matches = 0; 00508 00509 if (p->buf_size < 5) 00510 return 0; 00511 00512 state = AV_RB32(p->buf); 00513 for (i = 4; i < p->buf_size; i++) { 00514 if ((state & 0xffffff7f) == 0x1f07003f) 00515 matches++; 00516 // any section header, also with seq/chan num != 0, 00517 // should appear around every 12000 bytes, at least 10 per frame 00518 if ((state & 0xff07ff7f) == 0x1f07003f) 00519 secondary_matches++; 00520 if (state == 0x003f0700 || state == 0xff3f0700) 00521 marker_pos = i; 00522 if (state == 0xff3f0701 && i - marker_pos == 80) 00523 matches++; 00524 state = (state << 8) | p->buf[i]; 00525 } 00526 00527 if (matches && p->buf_size / matches < 1024*1024) { 00528 if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000)) 00529 return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match 00530 return AVPROBE_SCORE_MAX/4; 00531 } 00532 return 0; 00533 } 00534 00535 #if CONFIG_DV_DEMUXER 00536 AVInputFormat ff_dv_demuxer = { 00537 "dv", 00538 NULL_IF_CONFIG_SMALL("DV video format"), 00539 sizeof(RawDVContext), 00540 dv_probe, 00541 dv_read_header, 00542 dv_read_packet, 00543 dv_read_close, 00544 dv_read_seek, 00545 .extensions = "dv,dif", 00546 }; 00547 #endif