libavcodec/libtheoraenc.c

   1 /*
   2  * Copyright (c) 2006 Paul Richards <paul.richards@gmail.com>
   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  */
  20
  21 /*!
  22  * \file libtheoraenc.c
  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  */
  32
  33 /* FFmpeg includes */
  34 #include "libavutil/intreadwrite.h"
  35 #include "libavutil/log.h"
  36 #include "avcodec.h"
  37
  38 /* libtheora includes */
  39 #include <theora/theora.h>
  40
  41 typedef struct TheoraContext {
  42     theora_state t_state;
  43 } TheoraContext;
  44
  45 /*!
  46     Concatenates an ogg_packet into the extradata.
  47 */
  48 static int concatenate_packet(unsigned int* offset,
  49                               AVCodecContext* avc_context,
  50                               const ogg_packet* packet)
  51 {
  52     const char* message = NULL;
  53     uint8_t* newdata    = NULL;
  54     int newsize = avc_context->extradata_size + 2 + packet->bytes;
  55
  56     if (packet->bytes < 0) {
  57         message = "ogg_packet has negative size";
  58     } else if (packet->bytes > 0xffff) {
  59         message = "ogg_packet is larger than 65535 bytes";
  60     } else if (newsize < avc_context->extradata_size) {
  61         message = "extradata_size would overflow";
  62     } else {
  63         newdata = av_realloc(avc_context->extradata, newsize);
  64         if (!newdata)
  65             message = "av_realloc failed";
  66     }
  67     if (message) {
  68         av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message);
  69         return -1;
  70     }
  71
  72     avc_context->extradata      = newdata;
  73     avc_context->extradata_size = newsize;
  74     AV_WB16(avc_context->extradata + (*offset), packet->bytes);
  75     *offset += 2;
  76     memcpy(avc_context->extradata + (*offset), packet->packet, packet->bytes);
  77     (*offset) += packet->bytes;
  78     return 0;
  79 }
  80
  81 static av_cold int encode_init(AVCodecContext* avc_context)
  82 {
  83     theora_info t_info;
  84     theora_comment t_comment;
  85     ogg_packet o_packet;
  86     unsigned int offset;
  87     TheoraContext *h = avc_context->priv_data;
  88
  89     /* Set up the theora_info struct */
  90     theora_info_init(&t_info);
  91     t_info.width        = FFALIGN(avc_context->width,  16);
  92     t_info.height       = FFALIGN(avc_context->height, 16);
  93     t_info.frame_width  = avc_context->width;
  94     t_info.frame_height = avc_context->height;
  95     t_info.offset_x     = 0;
  96     t_info.offset_y     = avc_context->height & 0xf;
  97     /* Swap numerator and denominator as time_base in AVCodecContext gives the
  98      * time period between frames, but theora_info needs the framerate.  */
  99     t_info.fps_numerator   = avc_context->time_base.den;
 100     t_info.fps_denominator = avc_context->time_base.num;
 101     if (avc_context->sample_aspect_ratio.num) {
 102         t_info.aspect_numerator   = avc_context->sample_aspect_ratio.num;
 103         t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;
 104     } else {
 105         t_info.aspect_numerator   = 1;
 106         t_info.aspect_denominator = 1;
 107     }
 108     t_info.colorspace               = OC_CS_UNSPECIFIED;
 109     t_info.pixelformat              = OC_PF_420;
 110     t_info.keyframe_frequency       = avc_context->gop_size;
 111     t_info.keyframe_frequency_force = avc_context->gop_size;
 112     t_info.keyframe_mindistance     = avc_context->keyint_min;
 113
 114     t_info.quick_p                      = 1;
 115     t_info.dropframes_p                 = 0;
 116     t_info.keyframe_auto_p              = 1;
 117     t_info.keyframe_data_target_bitrate = t_info.target_bitrate * 1.5;
 118     t_info.keyframe_auto_threshold      = 80;
 119     t_info.noise_sensitivity            = 1;
 120     t_info.sharpness                    = 0;
 121
 122     if (avc_context->flags & CODEC_FLAG_QSCALE) {
 123         /* to be constant with the libvorbis implementation, clip global_quality to 0 - 10
 124            Theora accepts a quality parameter p, which is:
 125                 * 0 <= p <=63
 126                 * an int value
 127          */
 128         t_info.quality        = av_clip(avc_context->global_quality / (float)FF_QP2LAMBDA, 0, 10) * 6.3;
 129         t_info.target_bitrate = 0;
 130     } else {
 131         t_info.target_bitrate = avc_context->bit_rate;
 132         t_info.quality        = 0;
 133     }
 134
 135     /* Now initialise libtheora */
 136     if (theora_encode_init(&(h->t_state), &t_info)) {
 137         av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n");
 138         return -1;
 139     }
 140
 141     /* Clear up theora_info struct */
 142     theora_info_clear(&t_info);
 143
 144     /*
 145         Output first header packet consisting of theora
 146         header, comment, and tables.
 147
 148         Each one is prefixed with a 16bit size, then they
 149         are concatenated together into ffmpeg's extradata.
 150     */
 151     offset = 0;
 152
 153     /* Header */
 154     theora_encode_header(&(h->t_state), &o_packet);
 155     if (concatenate_packet(&offset, avc_context, &o_packet))
 156         return -1;
 157
 158     /* Comment */
 159     theora_comment_init(&t_comment);
 160     theora_encode_comment(&t_comment, &o_packet);
 161     if (concatenate_packet(&offset, avc_context, &o_packet))
 162         return -1;
 163     /* Clear up theora_comment struct before we reset the packet */
 164     theora_comment_clear(&t_comment);
 165     /* And despite documentation to the contrary, theora_comment_clear
 166      * does not release the packet */
 167     ogg_packet_clear(&o_packet);
 168
 169     /* Tables */
 170     theora_encode_tables(&(h->t_state), &o_packet);
 171     if (concatenate_packet(&offset, avc_context, &o_packet))
 172         return -1;
 173
 174     /* Set up the output AVFrame */
 175     avc_context->coded_frame= avcodec_alloc_frame();
 176
 177     return 0;
 178 }
 179
 180 static int encode_frame(AVCodecContext* avc_context, uint8_t *outbuf,
 181                         int buf_size, void *data)
 182 {
 183     yuv_buffer t_yuv_buffer;
 184     TheoraContext *h = avc_context->priv_data;
 185     AVFrame *frame = data;
 186     ogg_packet o_packet;
 187     int result;
 188
 189     assert(avc_context->pix_fmt == PIX_FMT_YUV420P);
 190
 191     /* Copy planes to the theora yuv_buffer */
 192     if (frame->linesize[1] != frame->linesize[2]) {
 193         av_log(avc_context, AV_LOG_ERROR, "U and V stride differ\n");
 194         return -1;
 195     }
 196
 197     t_yuv_buffer.y_width   = FFALIGN(avc_context->width, 16);
 198     t_yuv_buffer.y_height  = FFALIGN(avc_context->height, 16);
 199     t_yuv_buffer.y_stride  = frame->linesize[0];
 200     t_yuv_buffer.uv_width  = t_yuv_buffer.y_width / 2;
 201     t_yuv_buffer.uv_height = t_yuv_buffer.y_height / 2;
 202     t_yuv_buffer.uv_stride = frame->linesize[1];
 203
 204     t_yuv_buffer.y = frame->data[0];
 205     t_yuv_buffer.u = frame->data[1];
 206     t_yuv_buffer.v = frame->data[2];
 207
 208     /* Now call into theora_encode_YUVin */
 209     result = theora_encode_YUVin(&(h->t_state), &t_yuv_buffer);
 210     if (result) {
 211         const char* message;
 212         switch (result) {
 213         case -1:
 214             message = "differing frame sizes";
 215             break;
 216         case OC_EINVAL:
 217             message = "encoder is not ready or is finished";
 218             break;
 219         default:
 220             message = "unknown reason";
 221             break;
 222         }
 223         av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result);
 224         return -1;
 225     }
 226
 227     /* Pick up returned ogg_packet */
 228     result = theora_encode_packetout(&(h->t_state), 0, &o_packet);
 229     switch (result) {
 230     case 0:
 231         /* No packet is ready */
 232         return 0;
 233     case 1:
 234         /* Success, we have a packet */
 235         break;
 236     default:
 237         av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result);
 238         return -1;
 239     }
 240
 241     /* Copy ogg_packet content out to buffer */
 242     if (buf_size < o_packet.bytes) {
 243         av_log(avc_context, AV_LOG_ERROR, "encoded frame too large\n");
 244         return -1;
 245     }
 246     memcpy(outbuf, o_packet.packet, o_packet.bytes);
 247
 248     // HACK: does not take codec delay into account (neither does the decoder though)
 249     avc_context->coded_frame->pts = frame->pts;
 250
 251     return o_packet.bytes;
 252 }
 253
 254 static av_cold int encode_close(AVCodecContext* avc_context)
 255 {
 256     ogg_packet o_packet;
 257     TheoraContext *h = avc_context->priv_data;
 258     int result;
 259     const char* message;
 260
 261     result = theora_encode_packetout(&(h->t_state), 1, &o_packet);
 262     theora_clear(&(h->t_state));
 263     av_freep(&avc_context->coded_frame);
 264     av_freep(&avc_context->extradata);
 265     avc_context->extradata_size = 0;
 266
 267     switch (result) {
 268     case 0:  /* No packet is ready */
 269     case -1: /* Encoding finished */
 270         return 0;
 271     case 1:
 272         /* We have a packet */
 273         message = "gave us a packet";
 274         break;
 275     default:
 276         message = "unknown reason";
 277         break;
 278     }
 279     av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed (%s) [%d]\n", message, result);
 280     return -1;
 281 }
 282
 283 static const enum PixelFormat supported_pixel_formats[] = { PIX_FMT_YUV420P, PIX_FMT_NONE };
 284
 285 /*! AVCodec struct exposed to libavcodec */
 286 AVCodec libtheora_encoder = {
 287     .name = "libtheora",
 288     .type = CODEC_TYPE_VIDEO,
 289     .id = CODEC_ID_THEORA,
 290     .priv_data_size = sizeof(TheoraContext),
 291     .init = encode_init,
 292     .close = encode_close,
 293     .encode = encode_frame,
 294     .pix_fmts = supported_pixel_formats,
 295     .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"),
 296 };