Recherche avancée

Sur d’autres sites (8172)

• threads : Perform the generic progress cleanup more carefully.

  11 février 2012, par Michael Niedermayer

  threads : Perform the generic progress cleanup more carefully.

• Turn image sequence into video with transparency

  29 janvier 2014, par Cody Hatch

  I've got what seems like it should be a really simple problem, but it's proving much harder than I expected. Here's the issue :

  I've got a fairly large image sequence consisting of numbered frames (output from Maya, for what its worth). The images are currently in Targa (.tga) format, but I could convert them to PNGs or other arbitrary format if that matters. The important thing is, they've got an alpha channel.

  What I want to do is programatically turn them into a video clip. The format doesn't really matter, but it needs to be lossless and have an alpha channel. Uncompressed video in a Quicktime container would probably be ideal.

  My initial thought was ffmpeg, but after wasting most of a day on it it seems it's got no support at all for alpha channels. Either I'm missing something, or the underlying libavcodec just doesn't do it.

  So, what's the right way here ? A command line tool like ffmpeg would be nice, but any solution that runs on Windows and could be called from a script would be fine.

  Note : Having an alpha chanel in your video isn't actually all that uncommon, and it's really useful if you want to composite it on top of another video clip or a still image. As far as I know uncompressed video, the Quicktime Animation codec, and the Sorenson Video 3 codec all support tranparency, and I've heard H.264 does as well. All we're really talking about is 32-bit color depth, and that's pretty widely supported ; both Quicktime .mov files and Windowss .avi files can handle it, and probably a lot more too.

  Quicktime Pro is more than happy to turn an image sequence into a 32-bit .mov file. Hit export, change color depth to "Millions of Colors+", select the Animation codec, crank the quality up to 100, and there you are - losslessly compressed video, with an alpha chanel, and it'll play back almost anywhere since the codec has been part of Quicktime since version 1.0. The problem is, Quicktime Pro doesn't have any sort of command-line interface (at least on Windows). ffmpeg supports encoding using the Quicktime Animation codec (which it calls qtrle), but it only supports a bit-depth of 24 bits.

  The issue isn't finding a video format that supports an alpha channel. Quicktime Animation would be ideal, but even uncompressed video should work. The problem is finding a tool that supports it.

• YUY2 image ==>>sws_scale ==>>x264_encoder_encode doesn't work in Windows

  8 décembre 2011, par shiju sasi

  I have a multi media app in Windows using x264 built using MSYS-MingW and ffmpeg Windows binaries. This works for most of the cameras which capture data in RGB24 and RGB32 formats in most of the OSes. But when I tested the app on a Windows 7 (64 bit OS) Sony Vaio Laptop which has an integrated webcam capturing in YUY2 format, the x264_enoder_encode crashes. The sws_scale to convert the YUY2 data to YUV420 Planar any way works fine here too and returns proper stride values. Please check the relevant code fragments that I have attached below.

       x264_param_apply_profile(&m_param, "baseline");
  
         m_pEncoder = x264_encoder_open(&m_param);
  
         x264_encoder_parameters(m_pEncoder,&m_param);
  
  
  
         m_encoderConvertCtx = sws_getContext(g_iWidth, g_iHeight, PIX_FMT_YUYV422, SCALE_WIDTH, SCALE_HEIGHT, PIX_FMT_YUV420P, SWS_BICUBIC, NULL, NULL, NULL);
  
  
  
         x264_picture_t m_pic_in, m_pic_out;   //X264 picture variables to get the X264 encoded picture out.
  
         x264_picture_init(&m_pic_in);
  
   m_pic_in.i_type = X264_CSP_I420;
  
  
  
   x264_nal_t*  m_nals;
  
  
  
  
  
   srcstride = g_iWidth * 2;  //For YUYV422 Packed
  
  
  
   AVFrame* pictIn;
  
   AVFrame* pictOut;
  
  
  
   int iInBytes = avpicture_get_size(PIX_FMT_YUV420P, SCALE_WIDTH, SCALE_HEIGHT); 
  
   uint8_t* outbuffer = (uint8_t*)av_malloc(iInBytes); 
  
   pictOut = avcodec_alloc_frame(); 
  
  
  
   avpicture_fill((AVPicture*)pictOut, outbuffer, PIX_FMT_YUV420P, SCALE_WIDTH, SCALE_HEIGHT); 
  
  
  
   sws_scale(m_encoderConvertCtx, &in_buf, &srcstride, 0, g_iHeight, pictOut->data, pictOut->linesize); //Scale from YUYV422 Packed to YUV420 Plane
  
  
  
  
  
  ///Code after Scale begins
  
  
  
  memcpy(m_pic_in.img.plane[0],pictOut->data[0],SCALE_WIDTH * SCALE_HEIGHT);
  
  memcpy(m_pic_in.img.plane[1],pictOut->data[1],SCALE_WIDTH * SCALE_HEIGHT/4);
  
  memcpy(m_pic_in.img.plane[2],pictOut->data[2],SCALE_WIDTH * SCALE_HEIGHT/4);
  
  m_pic_in.img.plane[3] = 0; 
  
  
  
  for(int iPlane = 0; iPlane < 3; iPlane++)
  
  {
  
   m_pic_in.img.i_stride[iPlane] = pictOut->linesize[iPlane];
  
  }
  
  m_pic_in.img.i_stride[3] = 0;
  
  
  
  int frame_size = x264_encoder_encode(m_pEncoder, &m_nals, &i_nals, &m_pic_in, &m_pic_out);

  Please help if possible, as this has been consuming a lot of time at my end. But I am not able to dig in to the library side for debugging. Any experienced hands are requested to assist.