SkeyeExPlayer(Windows)开发之框架讲解-摩杜云开发者社区

SkeyeExPlayer for Windows是基于ffmpeg进行开发的全功能播放器，开发过程中参考了很多开源的播放器，诸如vlc和ffplay等，其中最强大的莫过于vlc，但是鉴于vlc框架过于庞大而其中仍存在诸多问题而舍弃了，而其他的更倾向于演示demo，只能提供部分借鉴意义；故而，SkeyeExPlayer 一贯秉承Skeye系列小而精，接口简单功能强大的宗旨从新设计了一套框架，该套框架能适应多线程调用以及多个播放实例同时运行，和SkeyePlayer一样Skeye; 当然，在此也郑重的感谢各大开源播放器以及ffmpeg的作者的无私奉献。

SkeyeExPlayer分为三大模块：打开模块，获取流数据模块，解码模块和渲染模块，其中：

(1) 打开模块打开流模块很简单，教科书式的调用方法：

player->avformat_context = avformat_alloc_context();
  player->avformat_context->interrupt_callback.callback = interrupt_cb;
  player->avformat_context->interrupt_callback.opaque = player;

  // open input file
  AVDictionary *options = NULL;
  //av_dict_set(&options, "rtsp_transport", "udp", 0);
  if (avformat_open_input(&player->avformat_context, url, fmt, &options) != 0) 
  {
    goto error_handler;
  }

  // find stream info
  if (avformat_find_stream_info(player->avformat_context, NULL) < 0) 
  {
    goto error_handler;
  }

  // set current audio & video stream
  for (i=0,idx=-1,cur=-1; i<(int)player->avformat_context->nb_streams; i++) {
switch (type) {
  case AVMEDIA_TYPE_AUDIO:
    // get last codec context
    if (player->acodec_context) {
      lastctxt = player->acodec_context;
    }

    // get new acodec_context & astream_timebase
    player->acodec_context   = player->avformat_context->streams[idx]->codec;
    player->astream_timebase = player->avformat_context->streams[idx]->time_base;

    // reopen codec
    if (lastctxt) avcodec_close(lastctxt);
    decoder = avcodec_find_decoder(player->acodec_context->codec_id);
    if (decoder && avcodec_open2(player->acodec_context, decoder, NULL) == 0) {
      player->astream_index = idx;
    }
    else {
      av_log(NULL, AV_LOG_WARNING, "failed to find or open decoder for audio !\n");
      player->astream_index = -1;
    }
    break;

  case AVMEDIA_TYPE_VIDEO:
    // get last codec context
    if (player->vcodec_context) {
      lastctxt = player->vcodec_context;
    }

    // get new vcodec_context & vstream_timebase
    player->vcodec_context   = player->avformat_context->streams[idx]->codec;
    player->vstream_timebase = player->avformat_context->streams[idx]->time_base;

    // reopen codec
    if (lastctxt) avcodec_close(lastctxt);
    decoder = avcodec_find_decoder(player->vcodec_context->codec_id);
    if (decoder && avcodec_open2(player->vcodec_context, decoder, NULL) == 0) {
      player->vstream_index = idx;
    }
    else {
      av_log(NULL, AV_LOG_WARNING, "failed to find or open decoder for video !\n");
      player->vstream_index = -1;
    }
    break;

  case AVMEDIA_TYPE_SUBTITLE:
    return -1; // todo...
    }
  }
  if (idx == -1) return -1;
  // for audio
  if (player->astream_index != -1)
  {
    arate   = player->acodec_context->sample_rate;
    aformat = player->acodec_context->sample_fmt;
    alayout = player->acodec_context->channel_layout;
    //++ fix audio channel layout issue
    if (alayout == 0) {
      alayout = av_get_default_channel_layout(player->acodec_context->channels);
    }
    //-- fix audio channel layout issue
  }

  // for video
  if (player->vstream_index != -1) {
    vrate = player->avformat_context->streams[player->vstream_index]->r_frame_rate;
    if (vrate.num / vrate.den >= 100) {
      vrate.num = 25;
      vrate.den = 1;
    }
    player->vcodec_context->pix_fmt = vformat;
    width   = player->vcodec_context->width;
    height  = player->vcodec_context->height;
  }

首先，avformat_open_input打开一个流，为了避免在打开流的时候出现阻塞，我们创建一个线程来执行，同时，为了防止ffmpeg内部出现持久行的阻塞，我们传入阻塞回调函数，在关闭流或者其他必要的时候解除阻塞；avformat_find_stream_info获取流的解码信息，根据音视频以及字幕的解码信息初始化解码器；

(2) 读取流数据模块

retv = av_read_frame(player->avformat_context, packet);
    //++ play completed ++//
    if (retv < 0)
    {
      if (player->avformat_context->pb && player->avformat_context->pb->error)
      {   
        //告知播放实时流中断
        player->error_flag = 1;
        //创建断线重连错误检测线程
        //  [9/4/2017 swordtwelve]
        break;
      }
      player->player_status |= PS_D_PAUSE;
      pktqueue_write_post_i(player->pktqueue, packet);
      usleep(20*1000); 
      continue;
    }
    //-- play completed --//
    player->error_flag = 0;//-1=初始化 0=正常 1-n错误代码

    // audio
    if (packet->stream_index == player->astream_index)
    {
      pktqueue_write_post_a(player->pktqueue, packet);
    }

    // video
    if (packet->stream_index == player->vstream_index)
    {
      pktqueue_write_post_v(player->pktqueue, packet);
    }

    if (  packet->stream_index != player->astream_index
      && packet->stream_index != player->vstream_index )
    {
      av_packet_unref(packet); // free packet
      pktqueue_write_post_i(player->pktqueue, packet);
    }
  }

读取数据模块超级简单，创建一个线程循环执行av_read_frame，读取到一帧就将其放入队列，这里采用了ffplay的阻塞的方式来处理队列的消费者和生产者的问题，这块有待优化，后续将改成无锁循环队列模式，如SkeyePlayer。

(3) 解码模块解码模块分为音频和视频解码模块，音视频的解码流程非常相似，主要分为三步： a. 从队列中读取音视频编码数据； b. 音视频分别采用avcodec_decode_audio4和avcodec_decode_video2进行解码； c. 音视频渲染；这里着重讲解视频的解码后的过程，其中涉及到解码后的原始图像数据进行处理，解码出一帧图像以后，我们需要对其进行字幕和图像或者其他的视频图像的叠加，借助ffmpeg强大的图像转换和缩放能力，借助VFX库我们很容易实现：

consumed = avcodec_decode_video2(player->vcodec_context, vframe, &gotvideo, packet);
      if (consumed < 0) {
        av_log(NULL, AV_LOG_WARNING, "an error occurred during decoding video.\n");
        break;
      }

      if (gotvideo) 
      {
        // 解码视频帧添加特技处理 [9/7/2017 dingshuai]
        // 1. 叠加图片
        // 2. 叠加字母
        // 3. 画框...
        // 对解码帧进行特技处理（字符，图片叠加，添加特效） [Dingshuai 2017/08/07]
#if 1
        WaterMarkInfo g_waterMarkInfo = player->vfxConfigInfo.warkMarkInfo;
        if (g_waterMarkInfo.bIsUseWaterMark)
        {
          if (player->vcodec_context->width != vframe->width   || 
            player->vcodec_context->height != vframe->height || 
            player->vfxConfigInfo.warkMarkInfo.bResetWaterMark )
          {
            //初始化水印叠加
            //;表示台标位置：1 == 左上 2 == 右上 3 == 左下 4 == 右下
            //eWaterMarkPos = 3

            //;水印顶点x轴坐标，建议不小于0；不大于视频宽度
            //nLeftTopX = 0

            //;水印顶点y轴坐标，建议不小于0；不大于视频高度
            //nLeftTopY = 480

            //;水印风格：0 - 6
            //eWatermarkStyle = 3

            //;水印图像文件路径LOGO.png
            //strWMFilePath = .\Res\logo.png
            switch (g_waterMarkInfo.eWaterMarkPos)
            {
            case POS_LEFT_TOP:
              g_waterMarkInfo.nLeftTopX = 0;
              g_waterMarkInfo.nLeftTopY = 0;
              break;
            case POS_RIGHT_TOP:
              g_waterMarkInfo.nLeftTopX = vframe->width;
              g_waterMarkInfo.nLeftTopY = 0;
              break;
            case POS_LEFT_BOTTOM:
              g_waterMarkInfo.nLeftTopX = 0;
              g_waterMarkInfo.nLeftTopY = vframe->height;
              break;
            case POS_RIGHT_BOTTOM:
              g_waterMarkInfo.nLeftTopX = vframe->width;
              g_waterMarkInfo.nLeftTopY = vframe->height;
              break;
            }

            player->vfxHandle->SetVideoInVideoParam( 101, 0, 0, vframe->width,
              vframe->height, 100, 100, 100);

            player->vfxHandle->SetLogoImage(g_waterMarkInfo.strWMFilePath, g_waterMarkInfo.nLeftTopX,
              g_waterMarkInfo.nLeftTopY, g_waterMarkInfo.bIsUseWaterMark, g_waterMarkInfo.eWatermarkStyle);

            player->vfxConfigInfo.warkMarkInfo.bResetWaterMark = FALSE;
          }
        }


        //初始化字幕信息
        VideoTittleInfo tittleInfo = player->vfxConfigInfo.tittleInfo;
        if(tittleInfo.bResetTittleInfo)
        {

          //  -->1、初始化创建字幕指针,并初始化视频长宽参数    m_pVideoVfxMakerInfo->nDesWidth, m_pVideoVfxMakerInfo->nDesHeight,  m_pVideoVfxMakerInfo->strDesBytesType);
          player->vfxHandle->CreateOverlayTitle(vframe->width, vframe->height, ("YUY2"));

          //  -->2、设置字幕文字信息
          LOGFONTA inFont;
          inFont.lfHeight      = tittleInfo.nTittleHeight;
          inFont.lfWidth       = tittleInfo.nTittleWidth;
          inFont.lfEscapement  = 0;
          inFont.lfOrientation = 0;
          inFont.lfWeight      = tittleInfo.nFontWeight;//FW_NORMAL;
          inFont.lfItalic      = 0;
          inFont.lfUnderline   = 0;
          inFont.lfStrikeOut   = 0;
          inFont.lfCharSet        =GB2312_CHARSET;// ANSI_CHARSET;//134
          inFont.lfOutPrecision   =3;// OUT_DEFAULT_PRECIS;
          inFont.lfClipPrecision  = 2;//CLIP_DEFAULT_PRECIS;
          inFont.lfQuality        = 1;//PROOF_QUALITY;
          inFont.lfPitchAndFamily = 0;//49;//49

          strcpy(inFont.lfFaceName, tittleInfo.strFontType);//"华文新魏");//"华文隶书");"隶书"

          POINT pointTitle;

          if(tittleInfo.nMoveType==0)
          {
            pointTitle= tittleInfo.ptStartPosition;
            if(pointTitle.x<=0) pointTitle.x=1;
            if(pointTitle.x>=vframe->width) pointTitle.x=vframe->width/2;
          }
          else if(tittleInfo.nMoveType==1)//从左往右
          {

            pointTitle.x =  -1;
            pointTitle.y = tittleInfo.ptStartPosition.y;
          }
          else if(tittleInfo.nMoveType==2)
          {
            pointTitle.x =  vframe->width+1;
            pointTitle.y = tittleInfo.ptStartPosition.y;
          }

          player->vfxHandle->SetOverlayTitleInfo(tittleInfo.strTittleContent, 
            inFont, tittleInfo.nColorR, tittleInfo.nColorG,
            tittleInfo.nColorB, pointTitle);

          //-->3、设置字幕运行抓状态
          player->vfxHandle->SetOverlayTitleState(tittleInfo.nState);

          player->vfxConfigInfo.tittleInfo.bResetTittleInfo = FALSE;
        }

        if (player->vfxHandle && (g_waterMarkInfo.bIsUseWaterMark || tittleInfo.nState))//logo-水印 + 字幕 + ？？？
        {
          if (player->vcodec_context->width != vframe->width   || 
            player->vcodec_context->height != vframe->height )
          {
            if (pVfxBuffer)
            {
              free(pVfxBuffer);
              pVfxBuffer = NULL;
            }
          }

          int nBufSize = vframe->width*vframe->height << 1;
          if (!pVfxBuffer)
          {
            pVfxBuffer = (BYTE*)malloc(nBufSize); //缓存写入源数据 
            memset(pVfxBuffer, 0x00, nBufSize);
          }

          AVFrame src;
          av_image_fill_arrays(src.data, src.linesize, pVfxBuffer, outPixelFormat, vframe->width, vframe->height, 1);
          //YUV420 -> YUY2
          ConvertColorSpace(&src, outPixelFormat, vframe, inPixelFormat, vframe->width, vframe->height);
//          av_image_copy_to_buffer(pVfxBuffer, nBufSize,
//            vframe->data, vframe->linesize, AV_PIX_FMT_YUYV422,  vframe->width, vframe->height, 1);

          //水印叠加
          if(g_waterMarkInfo.bIsUseWaterMark)
            player->vfxHandle->AddWaterMask(pVfxBuffer);
          //OSD叠加
          if(tittleInfo.nState)
            player->vfxHandle->DoOverlayTitle(pVfxBuffer);

          //YUY2 -> I420
          //ConvertColorSpace(vframe, inPixelFormat, &src, outPixelFormat, vframe->width, vframe->height);
          av_image_fill_arrays(vframe->data, vframe->linesize, pVfxBuffer, outPixelFormat, vframe->width, vframe->height, 1);
          int nPixelFmt = AV_PIX_FMT_YUYV422;
          player_setparam(player, PARAM_RENDER_OUTFORMAT, &nPixelFmt); 
        }
        else
        {
          int nPixelFmt = AV_PIX_FMT_YUV420P;
          player_setparam(player, PARAM_RENDER_OUTFORMAT, &nPixelFmt); 
        }
#endif

由于视频渲染需要一定的时间，我们也将解码帧数据进入队列进行缓存，从而保证播放的流畅性；

(4) 渲染模块渲染模块分为音频渲染和视频渲染，音频渲染即播放，使用waveOutOpen，waveOutWrite等waveout函数即可实现，下面重点说一下视频渲染，视频渲染通俗讲也就是图像绘制，Windows平台可采用D3D,DDraw, GDI，OpenGL等多种方式进行呈现，本文主要采用3种渲染方式，D3D，GDI和OpenGL；为了保证渲染的流畅性，我们创建线程执行渲染， a. 获取解码图像队列； b. 音视频时间戳同步处理； c. D3D/gdi/openGL渲染：