Force Qt Camera video format

Question

I'm trying to use Qt Camera from QML.

I'm developing a custom VideoFilter:

QVideoFrame MyFilterRunnable::run(QVideoFrame* input, const QVideoSurfaceFormat&, RunFlags)

I started deploying the application on Windows and I have that:

• frame is mappable in QAbstractVideoBuffer::ReadWrite
• frame pixel format is PixelFormat::Format_BGR32

When I moved to Linux, unfortunately, everything changed, without changing the camera I have:

• The frame is only QAbstractVideoBuffer::ReadOnly
• frame pixel format is PixelFormat::Format_YUYV

And now I really don't know how to convert this frame to an OpenCV Mat .

Is there any way to choose which will be the pixel format of the Camera?

Answer 1

Right on Steven,

    if (input->pixelFormat() == QVideoFrame::Format_YUYV)
    {
        auto input_w = input->width ();
        auto input_h = input->height();

        auto cam_data = input->bits();

        cv::Mat yuyv(input_h, input_w,CV_8UC2, cam_data);
        cv::Mat rgb (input_h, input_w,CV_8UC3);

        cvtColor(yuyv, rgb, CV_YUV2BGR_YUYV);

        m_mat = rgb;
    }
    else
    if (input->pixelFormat() == QVideoFrame::Format_YUV420P || input->pixelFormat() == QVideoFrame::Format_NV12) {
        m_yuv = true;
        m_mat = yuvFrameToMat8(*input);
    } else {
        m_yuv = false;
        QImage wrapper = imageWrapper(*input);
        if (wrapper.isNull()) {
            if (input->handleType() == QAbstractVideoBuffer::NoHandle)
                input->unmap();
            return *input;
        }
        m_mat = imageToMat8(wrapper);
    }
    ensureC3(&m_mat);

Answer 2

I am facing the same problem between my Linux machine and a Raspberry : I am using the same camera and the the pixel formats gven by QVideoFrame are different. It has probably something to do with v4l2

About the conversion from YUYV to OpenCV Mat, this code worked for me :

QVideoFrame *input ;
...
auto input_w = input->width ();
auto input_h = input->height();

auto cam_data = input->bits();

cv::Mat yuyv(input_h, input_w,CV_8UC2, cam_data);
cv::Mat rgb (input_h, input_w,CV_8UC3);

cvtColor(yuyv, rgb, CV_YUV2BGR_YUYV);