Recherche avancée

Sur d’autres sites (12093)

• Scenecut detection and consistent GOP size - adaptive streaming

  24 mars 2014, par Tarun

  Sample Command :

  -map 0:0 -f mp4 -vcodec libx264 -preset slow -profile:v main -vf scale="640:trunc(ow/a/2)*2" -vb 700k -minrate 650k -maxrate 750k -bufsize 10000k -an -g 48 -x264opts keyint=48:min-keyint=10:scenecut=40 -flags +cgop -sc_threshold 40 -pix_fmt yuv420p -threads 0 -y

  There is as such no error in encoding, But I wanted to understand following points-

  1) the above command will ensure that range of GOP size is 10,48, and if any scene change value (frame1 to fame2) is >40%, then a keyframe is introduced there ?

  2) So that means in a 3 hours of source video, there is no guarantee that GOP size will remain same

  3) No consider, I am creating 7 MP4 files each with different bitrate and resolution. (These Mp4s will be encoded to smooth in later stage). Hence i am targeting adaptive streaming. But when I did that, I found that if GOP sizes were not consistent across each bitrates. What I mean by this is, for ex : if in Bitrate1 - GOP size is like 10, 20, 48 and so on, in other bitrates it wasnt in the same sequence. I hope my question makes sense.

  So is there a way to ensure that GOP size may vary across one single output. But it should be consistent across each bitrate provided that the source is same ?

  Also for adaptive streaming Is scenecut detection advisable ?

• OpenCV Java binds VideoCapture from file failing silently

  8 décembre 2014, par muz0

  I’m using OpenCV 2.4.8 with the supplied Windows 64bit Java jar. I’ve been making full use of OpenCV in my current environment up until this point.

  I’m unable to open video files using the
  VideoCapture class however webcam feeds work just fine.

  The below works as expected with video.isOpened returning true

      VideoCapture video = new VideoCapture();
  
      boolean result = video.open(0);

  The below fails with video.isOpened returning false

      VideoCapture video = new VideoCapture();
  
      boolean result = video.open("res/hand-test-1.mp4");

  Neither file formats seems to make a difference (These are converted, not just renamed in hope)

      video.open("res/hand-test-1.mp4");
  
      video.open("res/hand-test-1.avi");
  
      video.open("res/hand-test-1.wmv");

  Location seems to matter not either.

      video.open("C:/hand-test-1.mp4");
  
      video.open("C:\\hand-test-1.mp4");
  
      video.open("hand-test-1.mp4");

  Neither does garbage, no exception kicked up from OpenCV through Java either, seems to fail silently.

      video.open("ashdkfhkajsjdfkhaksdf");

  PATH contains the ffmpeg directory supplied with the opencv installation,

      C:\dev\opencv\sources\3rdparty\ffmpeg

  Right now I’ve run out of ideas, it seems like whatever I throw to the native via video.open(String) will return false.

  Any help would be much appreciated

• FFMPEG + SDL : How To Show Multi Frames In Separate Region ?

  19 mai 2014, par user3051473

  I’m using ffmpeg and SDL to develop an camera monitor APP and I want to show four separate streams at the same time(via four rtsp source). As following figure out :

  Now I achieve this by setting the SDL display region, and I identify different display region by using the different SDL_Rect identifying by the variable i(from 1 - 4).

  But this cause the efficient problem. For every region, I need to scale to the whole screen and lock/unlock screen, then display.

  I’m wondering that, Can I merge 4 different AVFrame(come from REG1 - REG4) into the whole picture and then show this picture ?

  Hope that the describe about is detail enough for you to konw my problem.
  Also thanks for your help.

  -------------------
  
  -        -        -
  
  -  REG1  -  REG2  -
  
  -        -        -
  
  -------------------
  
  -        -        -
  
  -  REG3  -  REG4  -
  
  -        -        -
  
  -------------------
  
  
  
  public void ShowFrame(AVFrame *pFrame, AVCodecContext *pCodecContext)
  
  {
  
      SDL_LockYUVOverlay(pBmp);
  
      AVPicture pict;
  
  
  
      // TODO scale pFrame to pict
  
      // ...
  
  
  
      SDL_UnlockYUVOverlay(pBmp);
  
  
  
      // WIDTH and HEIGHT represent the whole screen size
  
      SDL_Rect rect;
  
      rect.x = (WIDTH / 2) * (i % 2);
  
      rect.y = (HEIGHT / 2) * (i > 1 ? 1: 0);
  
      rect.w = WIDTH / 2;
  
      rect.h = HEIGHT / 2;
  
      SDL_DisplayYUVOverlay(pBmp, &rect);
  
  }