c++ - 在 RGB channel 中拆分连续缓冲区 (IMFMediaBuffer)

Question

我正在尝试将一个连续的缓冲区拆分为 3 个字节的 channel (RGB)。这是我让缓冲区充满图像的实际工作流程:

1. 设置源读取器 (MFVideoFormat_RGB32)
2. 接收视频格式信息
3. 读取第一张图片并转换为连续缓冲区...

除此之外，这是代码:

HRESULT hr = S_OK;
  IMFAttributes *attributes = NULL;
  SafeRelease(&_sourcereader);

  hr = MFCreateAttributes(&attributes, 1);
  if (FAILED(hr)) {
// TODO: set error
return false;
  }

  hr = attributes->SetUINT32(MF_SOURCE_READER_ENABLE_VIDEO_PROCESSING, true);
  if (FAILED(hr)) {
  // TODO: set error
  return false;
  }

  // conversion from qstring to const wchar*
  const WCHAR* wfilename = filename.toStdWString().c_str();

  // create source reader from file with attributes
  hr = MFCreateSourceReaderFromURL(wfilename, attributes, &_sourcereader);
  if (FAILED(hr)) {
      // TODO: set error
      return false;
  }

  // configure sourcereader for progressive RGB32 frames
  IMFMediaType *mediatype = NULL;
  hr = MFCreateMediaType(&mediatype);
  if (SUCCEEDED(hr))
  {
  hr = mediatype->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
  }

  if (SUCCEEDED(hr))
  {
  hr = mediatype->SetGUID(MF_MT_SUBTYPE, MFVideoFormat_RGB32);
  }

  if (SUCCEEDED(hr))
  {
  hr = _sourcereader->SetCurrentMediaType(
      (DWORD)MF_SOURCE_READER_FIRST_VIDEO_STREAM,
      NULL, mediatype);
  }

  // Ensure the stream is selected.
  if (SUCCEEDED(hr))
  {
  hr = _sourcereader->SetStreamSelection(
      (DWORD)MF_SOURCE_READER_FIRST_VIDEO_STREAM, TRUE);
  }

  if (FAILED(hr)) {
  // TODO: Error log for failed configuration
  std::cout << "(ConfigureSourceReader) Configuration failed" << std::endl;
  return false;
  }
  //------------------------------------------------------------------
  //---------------------- Get Video Format Infos --------------------
  //------------------------------------------------------------------
  GUID subtype = { 0 };
  // Get the media type from the stream.
  hr = _sourcereader->GetCurrentMediaType(
    (DWORD)MF_SOURCE_READER_FIRST_VIDEO_STREAM, &mediatype );

  // Make sure it is a video format.
  hr = mediatype->GetGUID(MF_MT_SUBTYPE, &subtype);
  if (subtype != MFVideoFormat_RGB32)
  {
  hr = E_UNEXPECTED;
  // TODO: Error log message
  SafeRelease(&mediatype);
  return false;
  }

  //------------------------------------------------------------------
  // Get the width and height
  UINT32  width = 0, height = 0;
  hr = MFGetAttributeSize(mediatype, MF_MT_FRAME_SIZE, &width, &height);
  if (FAILED(hr))
  {
  // TODO: Error log message
  SafeRelease(&mediatype);
  return false;
  }
  //assign dimensions to VideoInfo
  _videoinfo.imageHeight = height; _videoinfo.imageWidth = width;
  //std::cout << "(GetVideoFormat) width: " << width << ", height: " << height << std::endl;

  //------------------------------------------------------------------
  //get framerate
  UINT32  framerate_num = 0, framerate_denom = 0;
  hr = MFGetAttributeRatio(mediatype, MF_MT_FRAME_RATE, &framerate_num, &framerate_denom);
  if (FAILED(hr))
  {
  // TODO: Error log message
  SafeRelease(&mediatype);
  return false;
  }
  //set frame rate in struct
  _videoinfo.fps = framerate_num / framerate_denom; // TODO: check for valid fps 24,25,30 ...

  //------------------------------------------------------------------
  // Get length
  LONGLONG length = 0;
  PROPVARIANT var;
  PropVariantInit(&var);
  hr = _sourcereader->GetPresentationAttribute((DWORD)MF_SOURCE_READER_MEDIASOURCE,
MF_PD_DURATION,
&var
);
  if (SUCCEEDED(hr)) {
  assert(var.vt == VT_UI8);
  length = var.hVal.QuadPart;
  } else {
  // TODO : erro log msg
  return false;
  }

  //Get total framenumber and length: save to info struct
  _videoinfo.noofFrames = length / 10000000 * this->getFrameRate(); // incl. conversion from nano sec to sec
  _videoinfo.duration = length;

  //------------------------------------------------------------------
  // Get the stride to find out if the bitmap is top-down or bottom-up.
  LONG lStride = 0;
  lStride = (LONG)MFGetAttributeUINT32(mediatype, MF_MT_DEFAULT_STRIDE, 1);
  _videoinfo.stride = lStride;
  _videoinfo.bTopDown = (lStride > 0);
  //------------------------------------------------------------------

  SafeRelease(&mediatype);
  // return true and flag if initialization went well
  _bInitialized = true;
  return true;

之后我调用一个函数来读取单个帧(目前是第一帧)。

HRESULT hr = S_OK;
  IMFSample *pSample = NULL;
  IMFMediaBuffer *buffer = NULL;
  DWORD streamIndex, flags;
  LONGLONG llTimeStamp;
  // Read Sample (RGB32)
  hr = _sourcereader->ReadSample (
  (DWORD) MF_SOURCE_READER_FIRST_VIDEO_STREAM,
  0,
  &streamIndex,
  &flags,
  &llTimeStamp,
  &pSample);
  if (FAILED (hr)) {
  // TODO handle fail case
  }
  //convert sample data to buffer
  hr = pSample->ConvertToContiguousBuffer(&buffer);
  if (FAILED (hr)) {
      // TODO handle fail case
        }    

我知道通过调用函数 buffer->Lock(&pixels, NULL, &nPixels) 我可以获得存储在 pixels 中的字节流。在我的例子中，我创建了一个具有给定高度和宽度的自定义图像(来自 SourceReader；[第一个函数])。从空图像中我可以得到一个空的颜色矩阵，它必须用以下函数填充:Color (byte red, byte green, byte blue)

我不知道如何将我的 RGB32 BYTE 数组拆分成单个 channel 来填充我的图像？也许这是一个愚蠢的问题，但我对这个领域还比较陌生......

Answer 1

对于 RGB32，字节格式为:

R=红色 G=绿色 B=蓝色 A=透明度

RGBARGBARGBA...

下面显示了一个非常简单的提取 channel 的伪代码示例。

for (int row = 0; row < height; row++) {
    for (int col = 0; col < stride; col += 4) {
        redBuf[rIndex++] = sample[row * stride + col];
        greenBuf[gIndex++] = sample[row * stride + col + 1];
        blueBuf[bIndex++] = sample[row * stride + col + 2];
        transparencyBuf[tIndex++] = sample[row * stride + col + 3];
    }
}