Qt5 Separating a video widget - c++11

Question

I am trying to find a way to create and separate a video widget into two parts, in order to process stereo videos:
The first one would play a part of the video;
The second one would play the other part of the video.
I currently do not know where to start. I am searching around qt multimedia module, but I do not know how to achieve this behavior.
Does anyone have an idea?
I was also thinking to build two video widgets and run them into two threads but they have to be perfectly synchronized. The idea was to cut the video into two ones with ffmpeg and affecting each one to a video widget. However I do not think it would be easy to achieve this (each frame would have to be sync).
Thanks for your answers.

Answer 1

If your stereo video data is encoded in some special format that needs decoding on the codec/container format, I think that the QMultiMedia stuff in Qt is too basic for this kind of use case as it does not allow tuning into "one stream" of a multi-stream transport container.
However, if you have alternating scan-lines, alternating frames or even "side-by-side" or "over-and-under" image per frame encoded in a "normal" video stream, then all you will have to do is intercept the frames as they are being decoded and separate the frame into two QImages and display them.
That is definitely doable!
However depending on your video source and even the platform, you might want to select different methods. For example if you are using a QCamera as the source of your video you could use the QVideoProbe or QViewFinder approaches. Interrestingly the availability of those methods on different platforms vary, so definitely figure out that first.
If you are decoding video using QMediaPlayer, QVideoProbe will probably be the way to go.
For an inttroduction to how you can grab frames using the different methods, please look at some of the examples from the official documentation on the subject.
Here is a short example of using the QVideoProbe approach:
videoProbe = new QVideoProbe(this);
// Here, myVideoSource is a camera or other media object compatible with QVideoProbe
if (videoProbe->setSource(myVideoSource)) {
// Probing succeeded, videoProbe->isValid() should be true.
connect(videoProbe, SIGNAL(videoFrameProbed(QVideoFrame)),
this, SLOT(processIndividualFrame(QVideoFrame)));
}
// Cameras need to be started. Do whatever your video source requires to start here
myVideoSource->start();
// [...]
// This is the slot where the magic happens (separating each single frame from video into two `QImage`s and posting the result to two `QLabel`s for example):
void processIndividualFrame(QVideoFrame &frame){
QVideoFrame cloneFrame(frame);
cloneFrame.map(QAbstractVideoBuffer::ReadOnly);
const QImage image(cloneFrame.bits(),
cloneFrame.width(),
cloneFrame.height(),
QVideoFrame::imageFormatFromPixelFormat(cloneFrame.pixelFormat()));
cloneFrame.unmap();
QSize sz = image.size();
const int w = sz.width();
const int h2 = sz.height() / 2;
// Assumes "over-and-under" placement of stereo data for simplicity.
// If you instead need access to individual scanlines, please have a look at [this][2].
QImage leftImage = image.copy(0, 0, w, h2);
QImage rightImage = image.copy(0, h2, w, h2);
// Assumes you have a UI set up with labels named as below, and with sizing / layout set up correctly
ui->myLeftEyeLabel.setPixmap(QPixmap::fromImage(leftImage));
ui->myRightEyeLabel.setPixmap(QPixmap::fromImage(leftImage));
// Should play back rather smooth since they are effectively updated simultaneously
}
I hope this was useful.
BIG FAT WARNING: Only parts of this code has been tested or even compiled!