path: root/src/mov_encoder.cc

/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/***************************************************************************
 *            mov_encoder.cc
 *
 *  Sat Feb 19 14:13:19 CET 2005
 *  Copyright  2005 Bent Bisballe
 *  deva@aasimon.org
 ****************************************************************************/

/*
 * Originally from:
 * RTVideoRec Realtime video recoder and encoder for Linux
 *
 * Copyright (C) 2004  B. Stultiens
 * Copyright (C) 2004  Koen Otter and Glenn van der Meyden
 */

/*
 *    This file is part of MIaV.
 *
 *    MIaV is free software; you can redistribute it and/or modify
 *    it under the terms of the GNU General Public License as published by
 *    the Free Software Foundation; either version 2 of the License, or
 *    (at your option) any later version.
 *
 *    MIaV is distributed in the hope that it will be useful,
 *    but WITHOUT ANY WARRANTY; without even the implied warranty of
 *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *    GNU General Public License for more details.
 *
 *    You should have received a copy of the GNU General Public License
 *    along with MIaV; if not, write to the Free Software
 *    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA.
 */

/*
 * $Id$
 */

/*
 * $Log$
 * Revision 1.13  2005/05/07 10:25:34  deva
 *
 * Removed ffmpeg code from img_encoder and corrected decoding errors in mov_encoder
 *
 * Revision 1.12  2005/05/05 20:41:38  deva
 *
 * Removed the last pieces of ffmpeg... replaced it with libfame...
 * Not quite working yet, but all the major code is in place!
 *
 * Revision 1.11  2005/05/03 17:12:53  deva
 * Fixed a bug (forgot to set the frametime).
 *
 * Revision 1.10  2005/05/03 08:31:59  deva
 * Removed the error object, and replaced it with a more generic info object.
 *
 * Revision 1.9  2005/05/02 10:59:44  deva
 * Reverted to mpg file format (accidentally used avi with mpeg2)
 *
 * Revision 1.8  2005/05/01 09:56:26  deva
 * Added Id and Log tags to all files
 */

#include "mov_encoder.h"

#include "debug.h"
//av_alloc_format_context
//av_destruct_packet_nofree

MovEncoder::MovEncoder(const char *filename)
{
  // FIXME: Hmmm... should this be detected somewhere?!
  static int w = 720;
  static int h = 576;

  // Initialize yuv strucutre.
  yuv.w = w;
  yuv.h = h;
  yuv.p = w;
  yuv.y = new unsigned char [w*h];
  yuv.u = new unsigned char [w*h/4];
  yuv.v = new unsigned char [w*h/4];

  ////////////LIBDV STUFF///////////////
  
  dvdecoder = NULL; // Initialize in encode method

  /////////LIBFAME STUFF///////////

  // Allocate the output buffer.
  fame_buffer = new unsigned char [FAME_BUFFER_SIZE];

  // Open output file
  f=fopen(filename, "wb");
  // FIXME: check f == NULL

  // Open a new session of the fame library.
  fame_context = fame_open();
  // FIXME: check fame_context == NULL
  // (If initialization was successful, it returns a non-null context which 
  // can then be used for subsequent library calls.)

  /*
  typedef struct _fame_parameters_ {
    int width;                        // width of the video sequence
    int height;                       // height of the video sequence
    char const *coding;               // coding sequence
    int quality;                      // video quality
    int slices_per_frame;             // number of slices per frame
    unsigned int frames_per_sequence; // number of frames per sequence
    int frame_rate_num;               // numerator of frames per second
    int frame_rate_den;               // denominator of frames per second
    unsigned int shape_quality;       // binary shape quality
    unsigned int search_range;        // motion estimation search range
    unsigned char verbose;            // verbosity
  } fame_parameters_t;
  */
  // width and height specify the size of each frames of the video sequence. 
  // Both must be multiple of 16. width and height must be less than 4096x4096
  fame_par.width = 720;
  fame_par.height = 576;

  // coding is a string of I, P or B characters representing the sequence of 
  // frames the encoder must produce. I frames are intra-coded frames (similar 
  // to JPEG), whereas P and B frames are motion compressed, respectively 
  // predicted from past reference (I or P) frame, or bidirectionally predicted 
  // from past and future reference frame.
  static const char coding[] = "I\0";
  fame_par.coding = coding;

  // quality is a percentage, which controls compression versus quality.
  fame_par.quality = 100;

  // slices_per_frame is the number of frame slices per frame. More slices provide 
  // better error recovery. There must be at least one slice per frame, and at most 
  // height / 16
  fame_par.slices_per_frame = fame_par.height / 16;

  // frames_per_sequence is the maximum number of frames contained in a video 
  // sequence.
  fame_par.frames_per_sequence = 1024;//25 * 60 * 60 * 2; // 25 fps in two hours.

  // frame_rate_num/frame_rate_den specify the number of frames per second for 
  // playback.
  fame_par.frame_rate_num = 10;//25; // 25 / 1 fps = 25 fps
  fame_par.frame_rate_den = 30000;//1;

  // shape_quality is percentage determing the average binary shape accuracy in 
  // video with arbitrary shape.
  fame_par.shape_quality = 100;

  // search_range specifies the motion estimation search range in pixel unit. 
  // Small search ranges work best with slow motion videos, whereas larger search 
  // ranges are rather for fast motion videos.
  fame_par.search_range = 10; // FIXME: No idea what this should be!?

  // verbose when set to 1 outputs information on copyright, modules used and 
  // current frame on standard error.
  fame_par.verbose = 0;

  fame_init(fame_context, &fame_par, fame_buffer, FAME_BUFFER_SIZE);
  // FIXME: fame_init return a new context, or NULL if an error occurred.
}

MovEncoder::~MovEncoder()
{
  int written = fame_close(fame_context);
  fwrite(fame_buffer, written, 1, f);
  fclose(f);
  delete [] fame_buffer;
  delete [] yuv.y;
  delete [] yuv.u;
  delete [] yuv.v;
}

void MovEncoder::encode(Frame *dvframe)
{
  encode_video(dvframe);
  encode_audio(dvframe);
}


#define _CLAMP(x) (unsigned char) ((x)<0?0:(((x)>255)?255:(x)))

inline void rgb_to_yuv(unsigned char *rgb, unsigned char &y, unsigned char &u, unsigned char &v)
{
  y = _CLAMP( 0.257 * rgb[0] + 0.504 * rgb[1] + 0.098 * rgb[2] + 16);
  v = _CLAMP( 0.439 * rgb[0] - 0.368 * rgb[1] - 0.071 * rgb[2] + 128);
  u = _CLAMP(-0.148 * rgb[0] - 0.291 * rgb[1] + 0.439 * rgb[2] + 128);
}

void MovEncoder::encode_video(Frame *dvframe)
{
  // Decode DV Frame to YUV
  int w = 720;
  int h = 576;
  unsigned char rgb[720*576*4];

  unsigned char *pixels[3];
  int pitches[3];

  pixels[ 0 ] = rgb;
  pixels[ 1 ] = NULL;
  pixels[ 2 ] = NULL;

  pitches[ 0 ] = 720 * 3;
  pitches[ 1 ] = 0;
  pitches[ 2 ] = 0;

  if(!dvdecoder) {
    dvdecoder = dv_decoder_new(FALSE/*this value is unused*/, FALSE, FALSE);
    dvdecoder->quality = DV_QUALITY_BEST;

    dv_parse_header(dvdecoder, dvframe->data);
    //dv_parse_packs(decoder, frame->data); // Not needed anyway!
    
    dvdecoder->system = e_dv_system_625_50;  // PAL lines, PAL framerate
    dvdecoder->sampling = e_dv_sample_422;  // 4 bytes y, 2 bytes u, 2 bytes v
    dvdecoder->std = e_dv_std_iec_61834;
    dvdecoder->num_dif_seqs = 12;
  }

  dv_decode_full_frame(dvdecoder, 
                       dvframe->data, 
                       e_dv_color_rgb,
                       pixels,
                       pitches);

  /*
                       e_dv_color_yuv,
                       (uint8_t**) &yuv.y, //pixels,
                       (int*) &yuv.w);//pitches);
  */  
  // cvt rgb to yuv
  for (int y=0; y<h; y+=2)
    {
      for (int x=0; x<w; x+=2)
        {
          unsigned char vy[4],vu[4],vv[4];
          
          rgb_to_yuv(rgb + 3*((y+0)*w + (x+0)), vy[0],vu[0],vv[0]);
          rgb_to_yuv(rgb + 3*((y+0)*w + (x+1)), vy[1],vu[1],vv[1]);
          rgb_to_yuv(rgb + 3*((y+1)*w + (x+0)), vy[2],vu[2],vv[2]);
          rgb_to_yuv(rgb + 3*((y+1)*w + (x+1)), vy[3],vu[3],vv[3]);
          // Y
          yuv.y[(y+0)*w+(x+0)] = vy[0];
          yuv.y[(y+0)*w+(x+1)] = vy[1];
          yuv.y[(y+1)*w+(x+0)] = vy[2];
          yuv.y[(y+1)*w+(x+1)] = vy[3];
          // Cb
          yuv.u[y*w/4+x/2] = (vu[0]+vu[1]+vu[2]+vu[3])/4;
          // Cr
          yuv.v[y*w/4+x/2] = (vv[0]+vv[1]+vv[2]+vv[3])/4;
        }
    }

  // Encode YUV frame and write it to disk.
  fame_start_frame(fame_context, &yuv, 0);
  int written;
  while((written = fame_encode_slice(fame_context))) {
    fwrite(fame_buffer, written, 1, f);
  }
  fame_end_frame(fame_context,0);
}

void MovEncoder::encode_audio(Frame *dvframe)
{
  // TODO: Do some audio stuff here sometime!
}