Processing a video stream using Streaming Engines

I need to process two video streams, co-relate them and respond with actions in real-time. Can someone help me with the following

1. How can i use open Streaming Engines like Flink or Spark for the above ?
2. How do i achieve this without using something like an OpenCV library ? By bringing in a library in between, the overall performance gets affected.
3. Any other suggestions for the above ?

Thanks a lot for the help.