Recherche avancée

Sur d’autres sites (8921)

• ffmpeg Bmp to yuv : Crash at sws_scale

  28 décembre 2015, par the-owner

  The context :
  I have a succession of continuous bitmap and I want to encode them into a light video format.
  I use ffmpeg version 2.8.3 (the build here), under qt5, qt IDE, and msvc2013 for win32.

  The problem :
  My code crash at sws_scale () (and sometimes at avcodec_encode_video2()). When I explore the stack, the crash event occurs at sws_getCachedContext (). (I can only see the stack with these ffmpeg builds).
  I only use these ffmpeg libraries (from the Qt .pro file) :

  LIBS += -lavcodec -lavformat -lswscale -lavutil

  It’s swscale which bug. And this is the code :

  void newVideo ()
  
  {
  
      ULONG_PTR gdiplusToken;
  
      GdiplusStartupInput gdiplusStartupInput;
  
      GdiplusStartup(&gdiplusToken, &gdiplusStartupInput, NULL);
  
  
  
      initBitmap (); //init bmp
  
      int screenWidth =  bmp.bmiHeader.biWidth;
  
      int screenHeight = bmp.bmiHeader.biHeight;
  
  
  
      AVCodec * codec;
  
      AVCodecContext * c = NULL;
  
      uint8_t * outbuf;
  
      int i, out_size, outbuf_size;
  
  
  
  
  
      avcodec_register_all();
  
  
  
      qDebug () << "Video encoding\n";
  
  
  
      // Find the mpeg1 video encoder
  
      codec = avcodec_find_encoder(AV_CODEC_ID_H264);
  
      if (!codec)
  
      {
  
          qDebug () << "Codec not found\n";
  
          avcodec_close(c);
  
          av_free(c);
  
          return;
  
      }
  
      else
  
          qDebug () << "H264 codec found\n";
  
  
  
      c = avcodec_alloc_context3(codec);
  
  
  
      c->bit_rate = 1000000;
  
      c->width = 800; // resolution must be a multiple of two (1280x720),(1900x1080),(720x480)
  
      c->height = 600;
  
      c->time_base.num = 1; // framerate numerator
  
      c->time_base.den = 25; // framerate denominator
  
      c->gop_size = 30; // emit one intra frame every ten frames
  
      c->max_b_frames = 1; // maximum number of b-frames between non b-frames
  
      c->pix_fmt = AV_PIX_FMT_YUV420P; //Converstion RGB to YUV ?
  
      c->codec_id = AV_CODEC_ID_H264;
  
  
  
      struct SwsContext* fooContext = sws_getContext(screenWidth, screenHeight,
  
                                                     AV_PIX_FMT_RGB32,
  
                                                     c->width, c->height,
  
                                                     AV_PIX_FMT_YUV420P,
  
                                                     SWS_FAST_BILINEAR,
  
                                                     NULL, NULL, NULL);
  
  
  
      // Open the encoder
  
      if (avcodec_open2(c, codec, NULL) < 0)
  
      {
  
          qDebug () << "Could not open codec\n";
  
          avcodec_close(c);
  
          av_free(c);
  
          return;
  
      }
  
      else qDebug () << "H264 codec opened\n";
  
  
  
      outbuf_size = 100000 + c->width*c->height*(32>>3);//*(32>>3); // alloc image and output buffer
  
      outbuf = static_cast(malloc(outbuf_size));
  
      qDebug() << "Setting buffer size to: " << outbuf_size << "\n";
  
  
  
      FILE* f = fopen("TEST.mpg","wb");
  
      if(!f) qDebug() << "x - Cannot open video file for writing\n";
  
      else qDebug() << "Opened video file for writing\n";
  
  
  
      // encode 5 seconds of video
  
      for (i = 0; i < STREAM_FRAME_RATE*STREAM_DURATION; i++) //the stop condition i < 5.0*5
  
      {
  
          qDebug () << "i = " << i;
  
          fflush(stdout);
  
  
  
          HBITMAP hBmp;
  
          if (GetScreen(hBmp) == -1) return;
  
          BYTE * pPixels;// = new BYTE [bmp.bmiHeader.biSizeImage];
  
          pPixels = getPixels (hBmp);
  
          DeleteObject (hBmp);
  
  
  
          int nbytes = avpicture_get_size(AV_PIX_FMT_YUV420P, c->width, c->height);
  
          uint8_t* outbuffer = (uint8_t*)av_malloc(nbytes*sizeof(uint8_t));
  
          if(!outbuffer) // check if(outbuf) instead
  
          {
  
              qDebug () << "Bytes cannot be allocated";
  
              return;
  
          }
  
  
  
          AVFrame* inpic = avcodec_alloc_frame(); //av_frame_alloc () ?
  
          AVFrame* outpic = avcodec_alloc_frame();
  
  
  
          outpic->pts = (int64_t)((float)i * (1000.0/((float)(c->time_base.den))) * 90);
  
          if (avpicture_fill((AVPicture*) inpic, (uint8_t*) pPixels, AV_PIX_FMT_RGB32,
  
                         screenWidth, screenHeight) < 0)
  
              qDebug () <<  "avpicture_fill Fill picture with image failed"; //Fill picture with image
  
  
  
          if(avpicture_fill((AVPicture*) outpic, outbuffer, AV_PIX_FMT_YUV420P,
  
                         c->width, c->height) < 0)
  
              qDebug () <<  "avpicture_fill failed";
  
  
  
          if (av_image_alloc(outpic->data, outpic->linesize, c->width, c->height,
  
                         c->pix_fmt, 1) < 0)
  
              qDebug () <<  "av_image_alloc failed";
  
  
  
          inpic->data[0] += inpic->linesize[0]*(screenHeight - 1); // Flipping frame
  
          inpic->linesize[0] = -inpic->linesize[0]; // Flipping frame
  
  
  
  ////////////////////////////HERE THE BUG////////////////////////////////
  
          sws_scale(fooContext,
  
                    inpic->data, inpic->linesize,
  
                    0, c->height,
  
                    outpic->data, outpic->linesize); //HERE THE BUG
  
  
  
          av_free_packet((AVPacket *)outbuf);
  
          // encode the image
  
          out_size = avcodec_encode_video2 (c, (AVPacket *) outbuf,
  
                                            (AVFrame *) outbuf_size, (int *) outpic);
  
  ///////////////////////THE CODE DONT GO BEYOND/////////////////////////////////
  
  
  
          qDebug () << "Encoding frame" << i <<" (size=" << out_size <<"\n";
  
          fwrite(outbuf, 1, out_size, f);
  
          delete [] pPixels;
  
          av_free(outbuffer);
  
          av_free(inpic);
  
          av_freep(outpic);
  
      }
  
  
  
      // get the delayed frames
  
      for(; out_size; i++)
  
      {
  
          fflush(stdout);
  
          out_size = avcodec_encode_video2 (c, (AVPacket *) outbuf,
  
                                            (AVFrame *) outbuf_size, NULL);
  
          qDebug () << "Writing frame" << i <<" (size=" << out_size <<"\n";
  
          fwrite(outbuf, 1, out_size, f);
  
      }
  
  
  
      // add sequence end code to have a real mpeg file
  
      outbuf[0] = 0x00;
  
      outbuf[1] = 0x00;
  
      outbuf[2] = 0x01;
  
      outbuf[3] = 0xb7;
  
      fwrite(outbuf, 1, 4, f);
  
      fclose(f);
  
  
  
      avcodec_close(c);
  
      free(outbuf);
  
      av_free(c);
  
      qDebug () << "Closed codec and Freed\n";
  
  }

  And the output :

  Video encoding
  
  
  
  H264 codec found
  
  
  
  H264 codec opened
  
  
  
  Setting buffer size to:  2020000 
  
  
  
  Opened video file for writing
  
  
  
  i =  0
  
  **CRASH**

  I have thougth that my bitmap wasn’t good so I have crafted a bitmap just for testing, the code was :

      uint8_t* pPixels = new uint8_t[Width * 3 * Height];
  
      int x = 50;
  
      for(unsigned int i = 0; i < Width * 3 * Height; i = i + 3) // loop for generating color changing images
  
      {
  
          pPixels [i] = x % 255; //R
  
          pPixels [i + 1] = (x) % 255; //G
  
          pPixels [i + 2] = (255 - x) % 255; //B
  
      }

  However the crash continue. Perhaps, it might prove that it’s not the bitmap (pPixels) which has a problem.

  If anyone know, why I get this bug : Maybe don’t I set one parameter well ? Or one ffmpeg deprecated function ? etc.

  ---

  EDIT 1 27/12/15

  Thanks to Ronald S. Bultje The function sws_scale () does not crash with this code, however I get an error from it bad dst image pointers. My code :

  //DESTINATION FRAME            
  
  if (avpicture_alloc ((AVPicture*) dst_frame, AV_PIX_FMT_YUV420P, c->width, c->height) < 0)
  
              {
  
                  qDebug () <<  "# avpicture_alloc failed";
  
                  return;
  
              }
  
              if(avpicture_fill((AVPicture*) dst_frame, NULL, AV_PIX_FMT_YUV420P,
  
                             c->width, c->height) < 0)
  
                  qDebug () <<  "avpicture_fill failed";
  
              avcodec_align_dimensions2 (c, &c->width, &c->height, dst_frame->linesize);
  
  
  
  //SOURCE FRAME
  
              if (avpicture_fill((AVPicture*) src_frame, (uint8_t *) pPixels, AV_PIX_FMT_RGB32,
  
                                 tmp_screenWidth, tmp_screenHeight) < 0)
  
                  qDebug () <<  "# avpicture_fill Fill picture with image failed"; //Fill picture with image
  
              avcodec_align_dimensions2 (c, &tmp_screenWidth, &tmp_screenHeight, src_frame->linesize);
  
  
  
              struct SwsContext* conversionContext = sws_getContext(tmp_screenWidth,tmp_screenHeight,AV_PIX_FMT_RGB32,c->width, c->height, AV_PIX_FMT_YUV420P,SWS_FAST_BILINEAR, NULL, NULL, NULL);
  
  
  
              int output_Height = sws_scale(conversionContext,
  
                                            src_frame->data, src_frame->linesize,
  
                                            0, tmp_screenHeight,
  
                                            dst_frame->data, dst_frame->linesize); //return 0 -> bad dst image pointers error

  ---

  EDIT 2 28/12/15

  I have tried to follow the Ronald S. Bultje’s suggestion and now I get a bad src image pointers error, I have investigated and worked many hours but I do not find a solution. Here, there is the new snippet :

  AVFrame* src_frame = av_frame_alloc ();
  
  AVFrame* dst_frame = av_frame_alloc ();
  
  AVFrame* tmp_src_frame = av_frame_alloc ();
  
  
  
  /*........I do not use them until this snippet..........*/
  
  //DESTINATION
  
  //avpicture_free ((AVPicture*)dst_frame);
  
  avcodec_align_dimensions2 (c, &c->width, &c->height, dst_frame->linesize);
  
  if (avpicture_alloc ((AVPicture*) dst_frame, AV_PIX_FMT_YUV420P, c->width, c->height) < 0)
  
  {
  
      qDebug () <<  "# avpicture_alloc failed";
  
      return;
  
  }
  
  
  
  //SOURCE
  
  //stride = src_frame->linesize [0] = ((((screenWidth * bitPerPixel) + 31) & ~31) >> 3); do I need to do that ?
  
  //== stride - I have gotten this formula from : https://msdn.microsoft.com/en-us/library/windows/desktop/dd318229(v=vs.85).aspx
  
  if (avpicture_fill((AVPicture*) src_frame, (uint8_t *) pPixels, AV_PIX_FMT_RGB32,
  
                     screenWidth, screenHeight) < 0)
  
      qDebug () <<  "# avpicture_fill Fill picture with image failed"; //Fill picture with image
  
  //linesize [0] == 21760 like commented stride
  
  
  
  //Source TO TMP Source
  
  avcodec_align_dimensions2 (c, &tmp_screenWidth, &tmp_screenHeight, tmp_src_frame->linesize);
  
  if (avpicture_fill((AVPicture*) tmp_src_frame, NULL, AV_PIX_FMT_RGB32,
  
                     tmp_screenWidth, tmp_screenHeight) < 0)
  
      qDebug () <<  "# avpicture_fill Fill picture with image failed"; //Fill picture with image
  
  
  
  av_picture_copy ((AVPicture*) tmp_src_frame, (const AVPicture*) src_frame, AV_PIX_FMT_RGB32,
  
                   screenWidth, screenHeight);
  
  
  
  struct SwsContext* conversionContext = sws_getContext(tmp_screenWidth, tmp_screenHeight,
  
                                                        AV_PIX_FMT_RGB32,
  
                                                        c->width, c->height,
  
                                                        AV_PIX_FMT_YUV420P,
  
                                                        SWS_FAST_BILINEAR,
  
                                                        NULL, NULL, NULL);
  
  
  
  int output_Height = sws_scale(conversionContext,
  
                                tmp_src_frame->data, tmp_src_frame->linesize,
  
                                0, tmp_screenHeight,
  
                                dst_frame->data, dst_frame->linesize);
  
  //ffmpeg error = bad src image pointers
  
  // output_Height == 0

  EDIT 3

  For temp Picture I have done an avcode_align_dimension2() then a avpicture_alloc() for allocating memory and avpicture_fill() in order to fill the picture pointer. Below the updated code :

  //DESTINATION
  
  //avpicture_free ((AVPicture*)dst_frame);
  
  avcodec_align_dimensions2 (c, &c->width, &c->height, dst_frame->linesize);
  
  if (avpicture_alloc ((AVPicture*) dst_frame, AV_PIX_FMT_YUV420P, c->width, c->height) < 0)
  
  {
  
      qDebug () <<  "# avpicture_alloc failed";
  
      return;
  
  }
  
  
  
  //SOURCE
  
  //src_frame->linesize [0] = ((((screenWidth * bpp) + 31) & ~31) >> 3);
  
  //src_frame->linesize [0] = stride;
  
  if (avpicture_fill((AVPicture*) src_frame, (uint8_t *) pPixels, AV_PIX_FMT_RGB32,
  
                     screenWidth, screenHeight) < 0)
  
      qDebug () <<  "# avpicture_fill Fill picture with image failed"; //Fill picture with image
  
  
  
  //Source TO TMP Source
  
  avcodec_align_dimensions2 (c, &tmp_screenWidth, &tmp_screenHeight, tmp_src_frame->linesize);
  
  if (avpicture_alloc ((AVPicture*) tmp_src_frame, AV_PIX_FMT_RGB32, tmp_screenWidth, tmp_screenHeight) < 0)
  
  {
  
      qDebug () <<  "# avpicture_alloc failed";
  
      return;
  
  }
  
  int outbuf_size = tmp_screenWidth*tmp_screenHeight*4;// alloc image and output buffer
  
  outbuf = static_cast(malloc(outbuf_size));
  
  if (avpicture_fill((AVPicture*) tmp_src_frame, outbuf, AV_PIX_FMT_RGB32,
  
                     tmp_screenWidth, tmp_screenHeight) < 0)
  
      qDebug () <<  "# avpicture_fill Fill picture with image failed"; //Fill picture with image
  
  av_picture_copy ((AVPicture*) tmp_src_frame, (const AVPicture*) src_frame, AV_PIX_FMT_RGB32,
  
                   tmp_screenWidth, tmp_screenHeight);
  
  
  
  struct SwsContext* conversionContext = sws_getContext(tmp_screenWidth, tmp_screenHeight,
  
                                                        AV_PIX_FMT_RGB32,
  
                                                        c->width, c->height,
  
                                                        AV_PIX_FMT_YUV420P,
  
                                                        SWS_FAST_BILINEAR,
  
                                                        NULL, NULL, NULL);
  
  
  
  int output_Height = sws_scale(conversionContext,
  
                                tmp_src_frame->data, tmp_src_frame->linesize,
  
                                0, tmp_screenHeight,
  
                                dst_frame->data, dst_frame->linesize);

  The call stack is as follow : av_picture_copy() is called then av_image_copy() then _VEC_memcpy() then fastcopy_I() and crash ... The problem is not the dimensions (tmp_screenWidth/Height) ? (With av_picture_copy () could we copy a picture P1 with dim W1xH1 to a picture P2 with dimension W2xH2 ?)

  EDIT 4

  Crash at av_picture_copy() which call _aligned_malloc() then av_image_copy _VEC_memcpy() and fastcopy_I()

  //SOURCE
  
  if (avpicture_fill((AVPicture*) src_frame, (uint8_t *) pPixels, AV_PIX_FMT_RGB32,
  
                     screenWidth, screenHeight) < 0)
  
      qDebug () <<  "# avpicture_fill Fill picture with image failed"; //Fill picture with image
  
  
  
  //Source TO TMP Source
  
  avcodec_align_dimensions2 (c, &tmp_screenWidth, &tmp_screenHeight, tmp_src_frame->linesize);
  
  if (avpicture_alloc ((AVPicture*) tmp_src_frame, AV_PIX_FMT_RGB32, tmp_screenWidth, tmp_screenHeight) < 0)
  
  {
  
      qDebug () <<  "# avpicture_alloc failed";
  
      return;
  
  }
  
  av_picture_copy ((AVPicture*) tmp_src_frame, (const AVPicture*) src_frame, AV_PIX_FMT_RGB32,
  
                   tmp_screenWidth, tmp_screenHeight);

• Encode a MPEG-1 video using directshow

  4 septembre 2015, par Kayani

  I need to know whether it is possible to encode images to MPEG-1 video using DirectShow from Microsoft. I found in their documentation that it is possible to Convert a video to MPEG-1 but there is nothing about images. Images in my case are coming at runtime.
  Note : I was previously using FFMPEG to achieve this

• Programatically get non-overlapping images from MP4

  15 février 2013, par Carlos F

  My ultimate goal is to get meaningful snapshots from MP4 videos that are either 30 min or 1 hour long. "Meaningful" is a bit ambitious, so I have simplified my requirements.

  The image should be crisp - non-overlapping, and ideally not blurry. Initially, I thought getting a keyframe would work, but I had no idea that keyframes could have overlapping images embedded in them like this :

  Of course, some keyframe images look like this and those are much better :

  

  I was wondering if someone might have source code to :

  Take a sequence of say 10-15 continuous keyframes (jpg or png) and identify the best keyframe from all of them.

  This must happen entirely programaticaly. I found this paper : http://research.microsoft.com/pubs/68802/blur_determination_compressed.pdf

  and felt that I could "rank" a few images based on the above paper, but then I was dissuaded by this link : Extracting DCT coefficients from encoded images and video given that my source video is an MP4. Of course, this confuses me because the input into the system is just a sequence of jpg images.

  Another link that is interesting is :

  Detection of Blur in Images/Video sequences

  However, I am not sure if this will work for "overlapping" images.

  Any ideas ?