I have a module that can help you, as long as you install OpenCV and FFmpeg yourself on the client machine:
While most commonly used to retransmit RTSP networked webcams in a Vision-friendly format, the implementation also attempts to load OpenCV's java library into Ignition's scripting environment, and this works in Vision Clients, too.
If you go that route, I'd recommend displaying frames with the Paintable Canvas.
(Note that while the jython wrappers look like the python OpenCV bindings, they are really the java bindings. Numpy is not part of the kit.)