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 St, Fifth Floor, Boston, MA  02110-1301  USA
  19  */
  20
  21 /*!
  22  * \file theoraenc.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 "avcodec.h"
  35 #include "log.h"
  36
  37 /* libtheora includes */
  38 #include <theora/theora.h>
  39
  40 typedef struct TheoraContext{
  41     theora_state t_state;
  42 } TheoraContext;
  43
  44 /*!
  45     Concatenates an ogg_packet into the extradata.
  46 */
  47 static int concatenate_packet(unsigned int* offset, AVCodecContext* avc_context, const ogg_packet* packet)
  48 {
  49     char* message = NULL;
  50     uint8_t* newdata = NULL;
  51     int newsize = avc_context->extradata_size + 2 + packet->bytes;
  52
  53     if (packet->bytes < 0) {
  54         message = "ogg_packet has negative size";
  55     } else if (packet->bytes > 0xffff) {
  56         message = "ogg_packet is larger than 65535 bytes";
  57     } else if (newsize < avc_context->extradata_size) {
  58         message = "extradata_size would overflow";
  59     } else {
  60         newdata = av_realloc(avc_context->extradata, newsize);
  61         if (newdata == NULL) {
  62             message = "av_realloc failed";
  63         }
  64     }
  65     if (message != NULL) {
  66         av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message);
  67         return -1;
  68     }
  69
  70     avc_context->extradata = newdata;
  71     avc_context->extradata_size = newsize;
  72     avc_context->extradata[ (*offset)++ ] = packet->bytes >> 8;
  73     avc_context->extradata[ (*offset)++ ] = packet->bytes & 0xff;
  74     memcpy( avc_context->extradata + (*offset), packet->packet, packet->bytes );
  75     (*offset) += packet->bytes;
  76     return 0;
  77 }
  78
  79 static int encode_init(AVCodecContext* avc_context)
  80 {
  81     theora_info t_info;
  82     theora_comment t_comment;
  83     ogg_packet o_packet;
  84     unsigned int offset;
  85     TheoraContext *h = avc_context->priv_data;
  86
  87     /* Set up the theora_info struct */
  88     theora_info_init( &t_info );
  89     t_info.width = avc_context->width;
  90     t_info.height = avc_context->height;
  91     t_info.frame_width = avc_context->width;
  92     t_info.frame_height = avc_context->height;
  93     t_info.offset_x = 0;
  94     t_info.offset_y = 0;
  95     /* Swap numerator and denominator as time_base in AVCodecContext gives the
  96      * time period between frames, but theora_info needs the framerate.  */
  97     t_info.fps_numerator = avc_context->time_base.den;
  98     t_info.fps_denominator = avc_context->time_base.num;
  99     if (avc_context->sample_aspect_ratio.num != 0) {
 100         t_info.aspect_numerator = avc_context->sample_aspect_ratio.num;
 101         t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;
 102     } else {
 103         t_info.aspect_numerator = 1;
 104         t_info.aspect_denominator = 1;
 105     }
 106     t_info.colorspace = OC_CS_UNSPECIFIED;
 107     t_info.pixelformat = OC_PF_420;
 108     t_info.target_bitrate = avc_context->bit_rate;
 109     t_info.keyframe_frequency = avc_context->gop_size;
 110     t_info.keyframe_frequency_force = avc_context->gop_size;
 111     t_info.keyframe_mindistance = avc_context->keyint_min;
 112     t_info.quality = 0;
 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     /* Now initialise libtheora */
 123     if (theora_encode_init( &(h->t_state), &t_info ) != 0) {
 124         av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n");
 125         return -1;
 126     }
 127
 128     /* Clear up theora_info struct */
 129     theora_info_clear( &t_info );
 130
 131     /*
 132         Output first header packet consisting of theora
 133         header, comment, and tables.
 134
 135         Each one is prefixed with a 16bit size, then they
 136         are concatenated together into ffmpeg's extradata.
 137     */
 138     offset = 0;
 139
 140     /* Header */
 141     theora_encode_header( &(h->t_state), &o_packet );
 142     if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
 143         return -1;
 144     }
 145
 146     /* Comment */
 147     theora_comment_init( &t_comment );
 148     theora_encode_comment( &t_comment, &o_packet );
 149     if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
 150         return -1;
 151     }
 152
 153     /* Tables */
 154     theora_encode_tables( &(h->t_state), &o_packet );
 155     if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
 156         return -1;
 157     }
 158
 159     /* Clear up theora_comment struct */
 160     theora_comment_clear( &t_comment );
 161
 162     /* Set up the output AVFrame */
 163     avc_context->coded_frame= avcodec_alloc_frame();
 164
 165     return 0;
 166 }
 167
 168 static int encode_frame(
 169     AVCodecContext* avc_context,
 170     uint8_t *outbuf,
 171     int buf_size,
 172     void *data)
 173 {
 174     yuv_buffer t_yuv_buffer;
 175     TheoraContext *h = avc_context->priv_data;
 176     AVFrame *frame = data;
 177     ogg_packet o_packet;
 178     int result;
 179
 180     assert(avc_context->pix_fmt == PIX_FMT_YUV420P);
 181
 182     /* Copy planes to the theora yuv_buffer */
 183     if (frame->linesize[1] != frame->linesize[2]) {
 184         av_log(avc_context, AV_LOG_ERROR, "U and V stride differ\n");
 185         return -1;
 186     }
 187
 188     t_yuv_buffer.y_width = avc_context->width;
 189     t_yuv_buffer.y_height = avc_context->height;
 190     t_yuv_buffer.y_stride = frame->linesize[0];
 191     t_yuv_buffer.uv_width = t_yuv_buffer.y_width / 2;
 192     t_yuv_buffer.uv_height = t_yuv_buffer.y_height / 2;
 193     t_yuv_buffer.uv_stride = frame->linesize[1];
 194
 195     t_yuv_buffer.y = frame->data[0];
 196     t_yuv_buffer.u = frame->data[1];
 197     t_yuv_buffer.v = frame->data[2];
 198
 199     /* Now call into theora_encode_YUVin */
 200     result = theora_encode_YUVin( &(h->t_state), &t_yuv_buffer );
 201     if (result != 0) {
 202         const char* message;
 203         switch (result) {
 204             case -1:
 205                 message = "differing frame sizes";
 206                 break;
 207             case OC_EINVAL:
 208                 message = "encoder is not ready or is finished";
 209                 break;
 210             default:
 211                 message = "unknown reason";
 212                 break;
 213         }
 214         av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result);
 215         return -1;
 216     }
 217
 218     /* Pick up returned ogg_packet */
 219     result = theora_encode_packetout( &(h->t_state), 0, &o_packet );
 220     switch (result) {
 221         case 0:
 222             /* No packet is ready */
 223             return 0;
 224         case 1:
 225             /* Success, we have a packet */
 226             break;
 227         default:
 228             av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result);
 229             return -1;
 230     }
 231
 232     /* Copy ogg_packet content out to buffer */
 233     if (buf_size < o_packet.bytes) {
 234         av_log(avc_context, AV_LOG_ERROR, "encoded frame too large\n");
 235         return -1;
 236     }
 237     memcpy(outbuf, o_packet.packet, o_packet.bytes);
 238
 239     return o_packet.bytes;
 240 }
 241
 242 static int encode_close(AVCodecContext* avc_context)
 243 {
 244     ogg_packet o_packet;
 245     TheoraContext *h = avc_context->priv_data;
 246     int result;
 247     const char* message;
 248
 249     result = theora_encode_packetout( &(h->t_state), 1, &o_packet );
 250     theora_clear( &(h->t_state) );
 251     switch (result) {
 252         case 0:/* No packet is ready */
 253         case -1:/* Encoding finished */
 254             return 0;
 255         case 1:
 256             /* We have a packet */
 257             message = "gave us a packet";
 258             break;
 259         default:
 260             message = "unknown reason";
 261             break;
 262     }
 263     av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed (%s) [%d]\n", message, result);
 264     return -1;
 265 }
 266
 267 static const enum PixelFormat supported_pixel_formats[] = { PIX_FMT_YUV420P, -1 };
 268
 269 /*! AVCodec struct exposed to libavcodec */
 270 AVCodec libtheora_encoder =
 271 {
 272     .name = "libtheora",
 273     .type = CODEC_TYPE_VIDEO,
 274     .id = CODEC_ID_THEORA,
 275     .priv_data_size = sizeof(TheoraContext),
 276     .init = encode_init,
 277     .close = encode_close,
 278     .encode = encode_frame,
 279     .pix_fmts = supported_pixel_formats,
 280 };