EMGUCV: VideoCapture with Hardware Acceleration using Direct GPU Frame

[Sarah635]

Hello,

I am using VideoCapture with hardware acceleration in a C# Windows Forms based application. The application is successfully invoking the nVidia decoder to decode 1080p Full HD video.

Now I have to perform certain processing on the captured video frames from VideoCapture API. Main steps of GPU processing are:

1. VideoCapture using Hardware Acceleration
2. Upload frame to GPU
3. Main processing on GPU
4. Download image back to CPU
5. Display the videoframe in an ImageBox UI control
6. Write the video frame to H.264 encoded file using hardware acceleration

My code is not running in real time and there is latency introduced in the video. I profiled the code and came to the point that maximum time is utilized by the Uploading and Downloading of frame between CPU and GPU.

Is there any possibility to minimize that GPU and CPU transfers ? Is it possible that in VideoCapture() call I will get GPU frame rather than CPU frame ??

Best Regards
SARAH