I am making a C# app to capture graphic from an avermedia pcie capture card. But it seems that there are no out of box tools to do so. So I made a C++ directshow app to do the capture, which is a console app and opens a capture window when running. How can I redirect the output to a C# app? for example, to a CaptureElement?
1 Answers
So you want to have XAML CaptureElement connected to AverMedia PCIe capture card. This sounds like a well-understood challenge overall, however every other piece of technology you mentioned is eventually a bad choice: DirectShow, multiple apps with piping, redirection and fitting of cutsom code to XAML CaptureElement control.
Microsoft has intentionally been limited ways you can integrate different APIs and so there are not so many ways to get everything together.
Let us go over the supposed integration path. The capture card is supposed to be shipped with a compatible driver:
Video capture devices are supported through the UVC class driver and must be compatible with UVC 1.1
When this is the case, such devices are visible to Media Foundation API handling video capture among tasks. XAML CaptureElement would be able to see a video capture device through this API and this way everything is supposed to work without need to fit anything from your end.
If this is not happening, it suggests you are dealing with an unsupported device coming without suitable or compatible driver.
Previous media API in Windows was DirectShow but its days are gone. It remains perfectly working as a legacy framework, a lot of applications out there are still relying on it. Specifically it will not integrate with new technology like XAML and UWP. More to that, even Media Foundation itself, the current media API, in its public offering is lagging behind when it comes to fitting with most recent technology. Having said that it is a good idea to stay well clear of DirectShow here if this is at all possible.
I see no need for cross-process design with video travelling between process through piping. There is no good reason for such design and even though this can work efficiently (Windows itself proves it can work great in terms of performance by having so called Frame Server service in it), this is not to be built on piping. In your case it is unlikely to be have to be built on multiple processes either. Instead you can develop a native code DLL project that takes care of video acquisition and connects to managed code via suitable glue layer: C++/CLI, COM, C++/WinRT and such.
Then next thing is fitting to XAML CaptureElement. The control is designed to work with Windows.Media.Capture.MediaCapture class that talks to hardware and you don't have suitable hardware as you plan to implement your own acquisition layer. Long story short you are not supposed to forward external data to CaptureElement and you would have hard time doing this. Your best strategy is to upload externally obtained data to Windows.Graphics.Imaging.SoftwareBitmap or alike and take involved performance impact as acceptable. That is, you will be dealing with video frames as images.
An alternative way is to upload acquired video frames into Direct 3D 11 textures and it would open you a more performant way of integration with video related controls, such as Windows.UI.Xaml.Controls.SwapChainPanel however it would also require that you put much more development effort in there.
Related