HW deinterlacing

[vlada]

Hi,
could someone please explain me how HW deinterlacing works? My idea is this: Video decoder passes RAW data to video renderer together with information if the video is interlaced or not. The video renderer then performs HW deinterlacing if needed.

Is my idea correct or not?

I'm asking this because I tried to play a MPEG-4 AVC video through ffdshow and MPC. In VMR9 my CPU was completely used, with overlay only a half CPU was utilized. I didn't see any difference in picture. Why is there such a big difference in CPU use between VMR and overlay?

[zambelli]

The decoder needs to signal to the VMR that the video is interlaced. It must support DirectShow's VideoInfoHeader2 interface in order to do this. DXVA doesn't need to be used for decoding in order for HW deinterlacing to work. The video could be decoded entirely in software.

The CPU usage between VMR and overlay has nothing to do with HW deinterlacing but with the nature of video overlay itself.

[vlada]

What exactly has impact on the CPU usage in VMR? Would a better graphics card improve it?

[zambelli]

I'm not sure what in VMR uses *more* CPU, but I do know that it typically doesn't write directly to dedicated video memory like overlay does. http://en.wikipedia.org/wiki/Video_overlay