Recherche avancée

Sur d’autres sites (2981)

• Use ffmpeg to match an image to source frames in video [closed]

  25 mai 2024, par user22335954

  I'm trying to write an application to split a single video into multiple pieces based on the appearance of a specific image. (Think title cards). I have video files that may have more than one episode or content inside of a single file and I want them split anywhere I find that title card or image.

  


  My application works by the user providing a timestamp in the format of 00:00:00 to specify the title card image which is then used like this :

  


  ffmpeg -i FILE -qmin 1 -qscale:v 1 -vframes 00:00:00 -f image2 img.png


  


  Now I want to compare that image (img.png) to the source video file using the following example command I've found :

  


  ffmpeg -i FILE -loop 1 -i img.png -an -filter_complex "blend=difference:shortest=1,blackframe=90:20" -f null


  


  I've had to play around with the blackframe=90:20 values to get what I think are correct matches, but I don't understand what these values and/or the blackframe filter is actually controlling. The blend documentation : https://ffmpeg.org/ffmpeg-filters.html#Examples-46 doesn't seem to go into much detail about what is actually happening. I do understand the difference blend means I'm essentially looking for the smallest difference, indicating a frame match to my img, but beyond that I'm sort of just guessing.

  


  Additionally, the output shows a bunch of :

  


  [Parsed_blackframe_1 @ 0x5c1183081880] frame:195 pblack:99 pts:6506 t:6.506000 type:B last_keyframe:135


  


  Based on the frames I can parse those out to find the non-sequential frames and find how how many segments I expect in the video, but when I go to split them, I don't know how to translate the frame or the t value into a timestamp format of 00:00:00. Even for matches that I'm 100% sure of, the frame values don't seem to line up with what I expect. For example, from watching the video, I know that a perfect match occurs at exactly 00:01:45, but the blackframe data says the match occurs at frame 1471 or t:49.08 (the video has a framerate of 29.97). 1471 / 29.97 is indeed 49.08, but that does not correlate to the actual time of 1:45 (105 seconds). How can I convert these values into timestamps (or just show the timestamps of the frames) ?

  


• LIVE555 RTSP H.264 Raw Video File Stream - ffplay Errors

  22 octobre 2015, par Chris.

  I am streaming a raw .h264 video file via RTSP using LIVE555.

  To receive the stream I am using ffplay. However, when watching the video I notice bad video quality and a bunch of errors in the ffplay-console :

  Input #0, rtsp, from 'rtsp://xx.xx.xxx.x/stream': sq=    0B f=0/0
  
    Metadata:
  
      title           : stream
  
      comment         : stream
  
  Duration: N/A, start: 0.099989, bitrate: N/A
  
      Stream #0:0: Video: h264 (High), yuv420p(tv, smpte170m/smpte170m/bt470m), 16
  
  80x1050 [SAR 1:1 DAR 8:5], 60 fps, 60 tbr, 90k tbn, 120 tbc
  
  [h264 @ 03f92100] RTP: missed 46 packetsq=   28KB sq=    0B f=1/1
  
  [h264 @ 03f92100] RTP: missed 74 packetsq=   23KB sq=    0B f=1/1
  
  [h264 @ 03f92100] RTP: missed 43 packets
  
  [h264 @ 03f92100] RTP: missed 35 packetsq=  179KB sq=    0B f=1/1
  
  [h264 @ 05710640] left block unavailable for requested intra4x4 mode -1 at 0 38
  
  [h264 @ 05710640] error while decoding MB 0 38, bytestream 48108
  
  [h264 @ 05710640] Cannot use next picture in error concealment
  
  [h264 @ 05710640] concealing 2989 DC, 2989 AC, 2989 MV errors in P frame
  
  [h264 @ 051043c0] left block unavailable for requested intra4x4 mode -1 at 0 26
  
  [h264 @ 051043c0] error while decoding MB 0 26, bytestream 5894
  
  [h264 @ 051043c0] concealing 4249 DC, 4249 AC, 4249 MV errors in I frame
  
  [h264 @ 03f92100] RTP: missed 68 packetsq=   28KB sq=    0B f=1/1
  
  [h264 @ 03f92100] RTP: missed 31 packetsq=  153KB sq=    0B f=1/1
  
  [h264 @ 052a0020] concealing 3292 DC, 3292 AC, 3292 MV errors in I frame
  
  [h264 @ 052a0020] Cannot use next picture in error concealment1/1
  
  [h264 @ 052a0020] concealing 2190 DC, 2190 AC, 2190 MV errors in P frame
  
  [h264 @ 03f92100] RTP: missed 69 packetsq=   27KB sq=    0B f=1/1
  
  [h264 @ 052a0020] concealing 3732 DC, 3732 AC, 3732 MV errors in I frame
  
  [h264 @ 03f92100] RTP: missed 26 packetsq=   30KB sq=    0B f=1/1
  
  ...

  How can I determine what’s wrong here ? Either with the stream or the file ?

• FFMPEG and DirectX Capture in C++

  13 décembre 2016, par tankyx

  I have a system that allows me to capture a window and save it as a mp4, using ffmpeg. I use gdigrab to capture the frame, but it is fairly slow (60ms per av_read_frame calls)

  I know I can capture a game using the DirectX API, but I don’t know how to convert the resulting BMP to an AVFrame.

  The following code is the DirectX code I use to capture the frame

  extern void* pBits;
  
  extern IDirect3DDevice9* g_pd3dDevice;
  
  IDirect3DSurface9* pSurface;
  
  g_pd3dDevice->CreateOffscreenPlainSurface(ScreenWidth, ScreenHeight,
  
                                        D3DFMT_A8R8G8B8, D3DPOOL_SCRATCH, 
  
                                        &pSurface, NULL);
  
  g_pd3dDevice->GetFrontBufferData(0, pSurface);
  
  D3DLOCKED_RECT lockedRect;
  
  pSurface->LockRect(&lockedRect,NULL,
  
                 D3DLOCK_NO_DIRTY_UPDATE|
  
                 D3DLOCK_NOSYSLOCK|D3DLOCK_READONLY)));
  
  for( int i=0 ; i < ScreenHeight ; i++)
  
  {
  
      memcpy( (BYTE*) pBits + i * ScreenWidth * BITSPERPIXEL / 8 , 
  
           (BYTE*) lockedRect.pBits + i* lockedRect.Pitch , 
  
           ScreenWidth * BITSPERPIXEL / 8);
  
  }
  
  g_pSurface->UnlockRect();
  
  pSurface->Release();

  And here is my read loop :

  {
  
      while (1) {
  
      if (av_read_frame(pFormatCtx, &packet) < 0 || exit)
  
          break;
  
      if (packet.stream_index == videoindex) {
  
          // Decode video frame
  
          av_packet_rescale_ts(&packet, { 1, std::stoi(pParser->GetVal("video-fps")) }, pCodecCtx->time_base);
  
          avcodec_decode_video2(pCodecCtx, pFrame, &frameFinished, &packet);
  
  
  
          if (frameFinished) {
  
              pFrame->pts = i;
  
              i++;
  
              sws_scale(sws_ctx, (uint8_t const * const *)pFrame->data, pFrame->linesize, 0, pCodecCtx->height, pFrameRGB->data, pFrameRGB->linesize);
  
              pFrameRGB->pts = pFrame->pts;
  
              enc.encodeFrame(pFrameRGB);
  
  
  
      }
  
      // Free the packet that was allocated by av_read_frame
  
      av_free_packet(&packet);
  
  }

  How can I create an AVFrame using the bmp I have, without using the av_read_frame ?