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, AVCodecContext* avc_context, const ogg_packet* packet)
  49 {
  50     const char* message = NULL;
  51     uint8_t* newdata = NULL;
  52     int newsize = avc_context->extradata_size + 2 + packet->bytes;
  53
  54     if (packet->bytes < 0) {
  55         message = "ogg_packet has negative size";
  56     } else if (packet->bytes > 0xffff) {
  57         message = "ogg_packet is larger than 65535 bytes";
  58     } else if (newsize < avc_context->extradata_size) {
  59         message = "extradata_size would overflow";
  60     } else {
  61         newdata = av_realloc(avc_context->extradata, newsize);
  62         if (newdata == NULL) {
  63             message = "av_realloc failed";
  64         }
  65     }
  66     if (message != NULL) {
  67         av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message);
  68         return -1;
  69     }
  70
  71     avc_context->extradata = newdata;
  72     avc_context->extradata_size = newsize;
  73     AV_WB16(avc_context->extradata + (*offset), packet->bytes);
  74     *offset += 2;
  75     memcpy( avc_context->extradata + (*offset), packet->packet, packet->bytes );
  76     (*offset) += packet->bytes;
  77     return 0;
  78 }
  79
  80 static av_cold int encode_init(AVCodecContext* avc_context)
  81 {
  82     theora_info t_info;
  83     theora_comment t_comment;
  84     ogg_packet o_packet;
  85     unsigned int offset;
  86     TheoraContext *h = avc_context->priv_data;
  87
  88     /* Set up the theora_info struct */
  89     theora_info_init( &t_info );
  90     t_info.width = FFALIGN(avc_context->width, 16);
  91     t_info.height = FFALIGN(avc_context->height, 16);
  92     t_info.frame_width = avc_context->width;
  93     t_info.frame_height = avc_context->height;
  94     t_info.offset_x = 0;
  95     t_info.offset_y = avc_context->height & 0xf;
  96     /* Swap numerator and denominator as time_base in AVCodecContext gives the
  97      * time period between frames, but theora_info needs the framerate.  */
  98     t_info.fps_numerator = avc_context->time_base.den;
  99     t_info.fps_denominator = avc_context->time_base.num;
 100     if (avc_context->sample_aspect_ratio.num != 0) {
 101         t_info.aspect_numerator = avc_context->sample_aspect_ratio.num;
 102         t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;
 103     } else {
 104         t_info.aspect_numerator = 1;
 105         t_info.aspect_denominator = 1;
 106     }
 107     t_info.colorspace = OC_CS_UNSPECIFIED;
 108     t_info.pixelformat = OC_PF_420;
 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
 113     t_info.quick_p = 1;
 114     t_info.dropframes_p = 0;
 115     t_info.keyframe_auto_p = 1;
 116     t_info.keyframe_data_target_bitrate = t_info.target_bitrate * 1.5;
 117     t_info.keyframe_auto_threshold = 80;
 118     t_info.noise_sensitivity = 1;
 119     t_info.sharpness = 0;
 120
 121     if (avc_context->flags & CODEC_FLAG_QSCALE) {
 122         /* to be constant with the libvorbis implementation, clip global_quality to 0 - 10
 123            Theora accepts a quality parameter p, which is:
 124                 * 0 <= p <=63
 125                 * an int value
 126          */
 127         t_info.quality = av_clip(avc_context->global_quality / (float)FF_QP2LAMBDA, 0, 10) * 6.3;
 128         t_info.target_bitrate = 0;
 129     } else {
 130         t_info.target_bitrate = avc_context->bit_rate;
 131         t_info.quality = 0;
 132     }
 133
 134     /* Now initialise libtheora */
 135     if (theora_encode_init( &(h->t_state), &t_info ) != 0) {
 136         av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n");
 137         return -1;
 138     }
 139
 140     /* Clear up theora_info struct */
 141     theora_info_clear( &t_info );
 142
 143     /*
 144         Output first header packet consisting of theora
 145         header, comment, and tables.
 146
 147         Each one is prefixed with a 16bit size, then they
 148         are concatenated together into ffmpeg's extradata.
 149     */
 150     offset = 0;
 151
 152     /* Header */
 153     theora_encode_header( &(h->t_state), &o_packet );
 154     if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
 155         return -1;
 156     }
 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 ) != 0) {
 162         return -1;
 163     }
 164     /* Clear up theora_comment struct before we reset the packet */
 165     theora_comment_clear( &t_comment );
 166     /* And despite documentation to the contrary, theora_comment_clear
 167      * does not release the packet */
 168     ogg_packet_clear(&o_packet);
 169
 170     /* Tables */
 171     theora_encode_tables( &(h->t_state), &o_packet );
 172     if (concatenate_packet( &offset, avc_context, &o_packet ) != 0) {
 173         return -1;
 174     }
 175
 176     /* Set up the output AVFrame */
 177     avc_context->coded_frame= avcodec_alloc_frame();
 178
 179     return 0;
 180 }
 181
 182 static int encode_frame(
 183     AVCodecContext* avc_context,
 184     uint8_t *outbuf,
 185     int buf_size,
 186     void *data)
 187 {
 188     yuv_buffer t_yuv_buffer;
 189     TheoraContext *h = avc_context->priv_data;
 190     AVFrame *frame = data;
 191     ogg_packet o_packet;
 192     int result;
 193
 194     assert(avc_context->pix_fmt == PIX_FMT_YUV420P);
 195
 196     /* Copy planes to the theora yuv_buffer */
 197     if (frame->linesize[1] != frame->linesize[2]) {
 198         av_log(avc_context, AV_LOG_ERROR, "U and V stride differ\n");
 199         return -1;
 200     }
 201
 202     t_yuv_buffer.y_width = FFALIGN(avc_context->width, 16);
 203     t_yuv_buffer.y_height = FFALIGN(avc_context->height, 16);
 204     t_yuv_buffer.y_stride = frame->linesize[0];
 205     t_yuv_buffer.uv_width = t_yuv_buffer.y_width / 2;
 206     t_yuv_buffer.uv_height = t_yuv_buffer.y_height / 2;
 207     t_yuv_buffer.uv_stride = frame->linesize[1];
 208
 209     t_yuv_buffer.y = frame->data[0];
 210     t_yuv_buffer.u = frame->data[1];
 211     t_yuv_buffer.v = frame->data[2];
 212
 213     /* Now call into theora_encode_YUVin */
 214     result = theora_encode_YUVin( &(h->t_state), &t_yuv_buffer );
 215     if (result != 0) {
 216         const char* message;
 217         switch (result) {
 218             case -1:
 219                 message = "differing frame sizes";
 220                 break;
 221             case OC_EINVAL:
 222                 message = "encoder is not ready or is finished";
 223                 break;
 224             default:
 225                 message = "unknown reason";
 226                 break;
 227         }
 228         av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result);
 229         return -1;
 230     }
 231
 232     /* Pick up returned ogg_packet */
 233     result = theora_encode_packetout( &(h->t_state), 0, &o_packet );
 234     switch (result) {
 235         case 0:
 236             /* No packet is ready */
 237             return 0;
 238         case 1:
 239             /* Success, we have a packet */
 240             break;
 241         default:
 242             av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result);
 243             return -1;
 244     }
 245
 246     /* Copy ogg_packet content out to buffer */
 247     if (buf_size < o_packet.bytes) {
 248         av_log(avc_context, AV_LOG_ERROR, "encoded frame too large\n");
 249         return -1;
 250     }
 251     memcpy(outbuf, o_packet.packet, o_packet.bytes);
 252
 253     // HACK: does not take codec delay into account (neither does the decoder though)
 254     avc_context->coded_frame->pts= frame->pts;
 255
 256     return o_packet.bytes;
 257 }
 258
 259 static av_cold int encode_close(AVCodecContext* avc_context)
 260 {
 261     ogg_packet o_packet;
 262     TheoraContext *h = avc_context->priv_data;
 263     int result;
 264     const char* message;
 265
 266     result = theora_encode_packetout( &(h->t_state), 1, &o_packet );
 267     theora_clear( &(h->t_state) );
 268     av_freep(&avc_context->coded_frame);
 269     av_freep(&avc_context->extradata);
 270     avc_context->extradata_size = 0;
 271
 272     switch (result) {
 273         case 0:/* No packet is ready */
 274         case -1:/* Encoding finished */
 275             return 0;
 276         case 1:
 277             /* We have a packet */
 278             message = "gave us a packet";
 279             break;
 280         default:
 281             message = "unknown reason";
 282             break;
 283     }
 284     av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed (%s) [%d]\n", message, result);
 285     return -1;
 286 }
 287
 288 static const enum PixelFormat supported_pixel_formats[] = { PIX_FMT_YUV420P, PIX_FMT_NONE };
 289
 290 /*! AVCodec struct exposed to libavcodec */
 291 AVCodec libtheora_encoder =
 292 {
 293     .name = "libtheora",
 294     .type = CODEC_TYPE_VIDEO,
 295     .id = CODEC_ID_THEORA,
 296     .priv_data_size = sizeof(TheoraContext),
 297     .init = encode_init,
 298     .close = encode_close,
 299     .encode = encode_frame,
 300     .pix_fmts = supported_pixel_formats,
 301     .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"),
 302 };