Recherche avancée

Sur d’autres sites (8423)

• Greed is Good ; Greed Works

  25 novembre 2010, par Multimedia Mike — VP8

  Greed, for lack of a better word, is good ; Greed works. Well, most of the time. Maybe.

  Picking Prediction Modes
  VP8 uses one of 4 prediction modes to predict a 16x16 luma block or 8x8 chroma block before processing it (for luma, a block can also be broken into 16 4x4 blocks for individual prediction using even more modes).

  So, how to pick the best predictor mode ? I had no idea when I started writing my VP8 encoder. I did not read any literature on the matter ; I just sat down and thought of a brute-force approach. According to the comments in my code :

  // naive, greedy algorithm :
  //   residual = source - predictor
  //   mean = mean(residual)
  //   residual -= mean
  //   find the max diff between the mean and the residual
  // the thinking is that, post-prediction, the best block will
  // be comprised of similar samples
  

  After removing the predictor from the macroblock, individual 4x4 subblocks are put through a forward DCT and quantized. Optimal compression in this scenario results when all samples are the same since only the DC coefficient will be non-zero. Failing that, when the input samples are at least similar to each other, few of the AC coefficients will be non-zero, which helps compression. When the samples are all over the scale, there aren’t a whole lot of non-zero coefficients unless you crank up the quantizer, which results in poor quality in the reconstructed subblocks.

  Thus, my goal was to pick a prediction mode that, when applied to the input block, resulted in a residual in which each element would feature the least deviation from the mean of the residual (relative to other prediction choices).

  Greedy Approach
  I realized that this algorithm falls into the broad general category of "greedy" algorithms— one that makes locally optimal decisions at each stage. There are most likely smarter algorithms. But this one was good enough for making an encoder that just barely works.

  Compression Results
  I checked the total file compression size on my usual 640x360 Big Buck Bunny logo image while forcing prediction modes vs. using my greedy prediction picking algorithm. In this very simple test, DC-only actually resulted in slightly better compression than the greedy algorithm (which says nothing about overall quality).

  prediction mode	quantizer index = 0 (minimum)	quantizer index = 10
  greedy	286260	98028
  DC	280593	95378
  vertical	297206	105316
  horizontal	295357	104185
  TrueMotion	311660	113480

  As another data point, in both quantizer cases, my greedy algorithm selected a healthy mix of prediction modes :

  • quantizer index 0 : DC = 521, VERT = 151, HORIZ = 183, TM = 65
  • quantizer index 10 : DC = 486, VERT = 167, HORIZ = 190, TM = 77

  Size vs. Quality
  Again, note that this ad-hoc test only measures one property (a highly objective one)— compression size. It did not account for quality which is a far more controversial topic that I have yet to wade into.

• Converting uint8_t data to AVFrame with FFmpeg

  30 octobre 2017, par J.Lefebvre

  I am currently working in C++ with the Autodesk 3DStudio Max 2014 SDK (toolset 100) and the Ffmpeg library in Visual Studio 2015 and trying to convert a DIB (Device Independent Bitmap) to uint8_t pointer array and then convert these data to an AVFrame.

  I don’t have any errors, but my video is still black and without meta data.
  (no time display, etc)

  I made approximatively the same with a Visual Studio Console application to convert jpeg image sequence from disk and this is working fine.
  (The only difference is that instead of converting jpeg to AVFrame with the Ffmpeg library, I try to convert raw data to an AVFrame.)

  So I think the problem could be either on the DIB conversion to the uint8_t data or the uint8_t data to the AVFrame.
  (The second is more plausible, because I used the SFML library to display a window with my rgb uint8_t* data for debuging and it is working fine.)

  I first initialize the ffmpeg library :

  This function is called once at the beginning.

  int Converter::Initialize(AVCodecID codec_id, int width, int height, int fps, const char *filename)
  
  {
  
      avcodec_register_all();
  
      av_register_all();
  
  
  
      AVCodec *codec;
  
      inputFrame = NULL;
  
      codecContext = NULL;
  
      pkt = NULL;
  
      file = NULL;
  
      outputFilename = new char[strlen(filename)]();
  
      *outputFilename = '\0';
  
      strcpy(outputFilename, filename);
  
  
  
      int ret;
  
  
  
      //Initializing AVCodecContext and getting PixelFormat supported by encoder
  
      codec = avcodec_find_encoder(codec_id);
  
      if (!codec)
  
          return 1;
  
  
  
      AVPixelFormat pixFormat = codec->pix_fmts[0];
  
      codecContext = avcodec_alloc_context3(codec);
  
      if (!codecContext)
  
          return 1;
  
  
  
      codecContext->bit_rate = 400000;
  
      codecContext->width = width;
  
      codecContext->height = height;
  
      codecContext->time_base.num = 1;
  
      codecContext->time_base.den = fps;
  
      codecContext->gop_size = 10;
  
      codecContext->max_b_frames = 1;
  
      codecContext->pix_fmt = pixFormat;
  
  
  
      if (codec_id == AV_CODEC_ID_H264)
  
          av_opt_set(codecContext->priv_data, "preset", "slow", 0);
  
  
  
      //Actually opening the encoder
  
      if (avcodec_open2(codecContext, codec, NULL) < 0)
  
          return 1;
  
  
  
      file = fopen(outputFilename, "wb");
  
      if (!file)
  
          return 1;
  
  
  
      inputFrame = av_frame_alloc();
  
      inputFrame->format = codecContext->pix_fmt;
  
      inputFrame->width = codecContext->width;
  
      inputFrame->height = codecContext->height;
  
  
  
      ret = av_image_alloc(inputFrame->data, inputFrame->linesize, codecContext->width, codecContext->height, codecContext->pix_fmt, 32);
  
  
  
      if (ret < 0)
  
          return 1;
  
  
  
      return 0;
  
  }

  Then for each frame, I get the DIB and convert to a uint8_t* it with this function :

  uint8_t* Util::ToUint8_t(RGBQUAD *data, int width, int height)
  
  {
  
      uint8_t* buf = (uint8_t*)data;
  
  
  
      int imageSize = width * height;
  
      size_t rgbquad_size = sizeof(RGBQUAD);
  
      size_t total_bytes = imageSize * rgbquad_size;
  
      uint8_t * pCopyBuffer = new uint8_t[total_bytes];
  
  
  
      for (int x = 0; x < width; x++)
  
      {
  
          for (int y = 0; y < height; y++)
  
          {
  
              int index = (x + width * y) * rgbquad_size;
  
              int invertIndex = (x + width* (height - y - 1)) * rgbquad_size;
  
  
  
              //BGRA to RGBA
  
              pCopyBuffer[index] = buf[invertIndex + 2];
  
              pCopyBuffer[index + 1] = buf[invertIndex + 1];
  
              pCopyBuffer[index + 2] = buf[invertIndex];
  
              pCopyBuffer[index + 3] = 0xFF;
  
          }
  
      }
  
  
  
      return pCopyBuffer;
  
  }
  
  
  
  void GetDIBBuffer(Interface* ip, BITMAPINFO *bmi, uint8_t** outBuffer)
  
  {
  
      int size;
  
  
  
      ViewExp& view = ip->GetActiveViewExp();
  
  
  
      view.getGW()->getDIB(NULL, &size);
  
  
  
      bmi = (BITMAPINFO *)malloc(size);
  
      BITMAPINFOHEADER *bmih = (BITMAPINFOHEADER *)bmi;
  
      view.getGW()->getDIB(bmi, &size);
  
  
  
      uint8_t * pCopyBuffer = Util::ToUint8_t(bmi->bmiColors, bmih->biWidth, bmih->biHeight);
  
  
  
      *outBuffer = pCopyBuffer;
  
  }

  This function is used to get the DIB :

  void GetViewportDIB(Interface* ip, BITMAPINFO *bmi, BITMAPINFOHEADER *bmih, BitmapInfo biFile, Bitmap *map)
  
  {
  
      int size;
  
  
  
      if (!biFile.Name()[0])
  
          return;
  
  
  
      ViewExp& view = ip->GetActiveViewExp();
  
  
  
      view.getGW()->getDIB(NULL, &size);
  
  
  
      bmi = (BITMAPINFO *)malloc(size);
  
      bmih = (BITMAPINFOHEADER *)bmi;
  
  
  
      view.getGW()->getDIB(bmi, &size);
  
  
  
      biFile.SetWidth((WORD)bmih->biWidth);
  
      biFile.SetHeight((WORD)bmih->biHeight);
  
      biFile.SetType(BMM_TRUE_32);
  
  
  
      map = TheManager->Create(&biFile);
  
      map->OpenOutput(&biFile);
  
      map->FromDib(bmi);
  
      map->Write(&biFile);
  
      map->Close(&biFile);
  
  }

  And after the conversion to AVFrame and video encoding :

  The EncodeFromMem function is call each frame.

  int Converter::EncodeFromMem(const char *outputDir, int frameNumber, uint8_t* data)
  
  {
  
      int ret;
  
  
  
      inputFrame->pts = frameNumber;
  
      EncodeFrame(data, codecContext, inputFrame, &pkt, file);
  
  
  
      return 0;
  
  }
  
  
  
  static void RgbToYuv(uint8_t *rgb, AVCodecContext *c, AVFrame *frame)
  
  {
  
      struct SwsContext *swsCtx = NULL;
  
      const int in_linesize[1] = { 3 * c->width };// RGB stride
  
      swsCtx = sws_getCachedContext(swsCtx, c->width, c->height, AV_PIX_FMT_RGB24, c->width, c->height, AV_PIX_FMT_YUV420P, 0, 0, 0, 0);
  
      sws_scale(swsCtx, (const uint8_t * const *)&rgb, in_linesize, 0, c->height, frame->data, frame->linesize);
  
  }
  
  
  
  static void EncodeFrame(uint8_t *rgb, AVCodecContext *c, AVFrame *frame, AVPacket **pkt, FILE *file)
  
  {
  
      int ret, got_output;
  
  
  
      RgbToYuv(rgb, c, frame);
  
  
  
      *pkt = av_packet_alloc();
  
      av_init_packet(*pkt);
  
      (*pkt)->data = NULL;
  
      (*pkt)->size = 0;
  
  
  
      ret = avcodec_encode_video2(c, *pkt, frame, &got_output);
  
      if (ret < 0)
  
      {
  
          fprintf(stderr, "Error encoding frame/n");
  
          exit(1);
  
      }
  
      if (got_output)
  
      {
  
          fwrite((*pkt)->data, 1, (*pkt)->size, file);
  
          av_packet_unref(*pkt);
  
      }
  
  }

  To finish I have a function that write the packets and free the memory :
  This function is called once at the end of the time range.

  int Converter::Finalize()
  
  {
  
      int ret, got_output;
  
      uint8_t endcode[] = { 0, 0, 1, 0xb7 };
  
  
  
      /* get the delayed frames */
  
      do
  
      {
  
          fflush(stdout);
  
          ret = avcodec_encode_video2(codecContext, pkt, NULL, &got_output);
  
          if (ret < 0)
  
          {
  
              fprintf(stderr, "Error encoding frame/n");
  
              return 1;
  
          }
  
          if (got_output)
  
          {
  
              fwrite(pkt->data, 1, pkt->size, file);
  
              av_packet_unref(pkt);
  
          }
  
      } while (got_output);
  
  
  
      fwrite(endcode, 1, sizeof(endcode), file);
  
      fclose(file);
  
  
  
      avcodec_close(codecContext);
  
      av_free(codecContext);
  
  
  
      av_frame_unref(inputFrame);
  
      av_frame_free(&inputFrame);
  
      //av_freep(&inputFrame->data[0]); //Crash
  
  
  
      delete outputFilename;
  
      outputFilename = 0;
  
  
  
      return 0;
  
  }

  EDIT :

  I modify my RgbToYuv function and create another one to convert back the yuv frame to an rgb one.

  This not really solve the problem, but maybe focus the problem on the conversion from YuvToRgb.

  This is the result of the conversion from YUV to RGB :

   ![YuvToRgb result] : https://img42.com/kHqpt+

  static void YuvToRgb(AVCodecContext *c, AVFrame *frame)
  
  {
  
      struct SwsContext *img_convert_ctx = sws_getContext(c->width, c->height, AV_PIX_FMT_YUV420P, c->width, c->height, AV_PIX_FMT_RGB24, SWS_BICUBIC, NULL, NULL, NULL);
  
      AVFrame * rgbPictInfo = av_frame_alloc();
  
      avpicture_fill((AVPicture*)rgbPictInfo, *(frame)->data, AV_PIX_FMT_RGB24, c->width, c->height);
  
      sws_scale(img_convert_ctx, frame->data, frame->linesize, 0, c->height, rgbPictInfo->data, rgbPictInfo->linesize);
  
  
  
      Util::DebugWindow(c->width, c->height, rgbPictInfo->data[0]);
  
  }
  
  static void RgbToYuv(uint8_t *rgb, AVCodecContext *c, AVFrame *frame)
  
  {
  
      AVFrame * rgbPictInfo = av_frame_alloc();
  
      avpicture_fill((AVPicture*)rgbPictInfo, rgb, AV_PIX_FMT_RGBA, c->width, c->height);
  
  
  
      struct SwsContext *swsCtx = sws_getContext(c->width, c->height, AV_PIX_FMT_RGBA, c->width, c->height, AV_PIX_FMT_YUV420P, SWS_BICUBIC, NULL, NULL, NULL);
  
      avpicture_fill((AVPicture*)frame, rgb, AV_PIX_FMT_YUV420P, c->width, c->height);    
  
      sws_scale(swsCtx, rgbPictInfo->data, rgbPictInfo->linesize, 0, c->height, frame->data, frame->linesize);
  
  
  
      YuvToRgb(c, frame);
  
  }

• How to compress video applying even blur effect

  17 octobre 2016, par Kukuster

  How to compress video applying well-looking apportioned blur effect, like in JPG image ?

  I tried some ffmpeg postprocesing libraries, they are fspp, spp, uspp (takes really long time to render), etc. I almost reached the goal using fspp with parameters 5:60:15 . But blur was stronger than needed, and it’s leave bad artifacts when i try to use less compression. Also uspp is does beautiful and compresses enough, but it’s leave about 50% of video unblured. I also haven’t much time to try all uspp features. Is there resolution special for this purpose ?

  The point is to implement video compression with the only side effect of compression approaches jpeg-compression-like blur or blur-mask-like. Also it would be very good if there is a simple option to choose between :
  1) more compress, less pretty blur / more strong blur ; and 2) less compress, prettier blur / less strong blur.

  I am used to use ffmpeg and i’m running linux, so it would be so nice if there is a way to solve this with ffmpeg.
  Here is my ffmpeg input data about the video streams :

    Duration : 00:01:03.02, start : 0.000000, bitrate : 4010 kb/s
      Stream #0:0(und) : Video : h264 (High) (avc1 / 0x31637661), yuv420p, 1280x720 [SAR 1:1 DAR 16:9], 3870 kb/s, 23.98 fps, 23.98 tbr, 24k tbn, 47.95 tbc (default)
      Metadata :
        handler_name : VideoHandler
      Stream #0:1(eng) : Audio : aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 133 kb/s (default)
      Metadata :
        handler_name : SoundHandler
  

  Edit : attaching jpeg pictured as example of desired output :