Hi mikenap,
Yes, you would have to use an HDMI/DVI receiver chip. Even of the FPGA could directly interface to the DVI signals, it would still need to do a lot of decoding since the DVI stream is much more than just the serialized video data. All HDMI receivers should be able to decode DVI. The output of such a chips is typically a BT1120 (or BT656 if SD) video stream which can be directly fed to the Clocked Video Input core in the VIP toolkit.
The FPGA would have to convert the BT1120 stream to an S-Video compatible stream. Afaik (could be wrong), S-Video is only defined for PAL/NTSC standard definition video (ie 422 13.5/27MHz pixel clock), while DVI outputs are often much higher resolutions and may have other frame rates as well. Converting the resolution can be done with a scaler from the VIP, but a frame rate conversion is maybe more tricky. A quick and dirty approach is to use the frame buffer with tripple buffering to drop/repeat frames as required. This may or may not be acceptable in your application since it may lead to "jerkiness".
On the output you need a video encoder which takes a BT656 stream and converts it to analog S-Video signals. There are many from AD/TI and others. The Clocked Video Output core provides a BT656 / BT1120 compatible output stream (with external or embedded syncs).
So, at least three chips then.
Regards,
Niki