PAL capture problems

[seaton]

I am using a fancy license plate camera, but finding weird results. From what I can tell the video must be PAL

Hardware:

LPR Camera:
PAL video 1V p-p 75Ω CVBS, colour burst suppressed for ANPR
752(H) x 582(V) effective picture elements, 680 TVL, Day/Night IR Cut, Colour/B&W

Capture devices:
Startech SVID2USB2. specs -> http://www.startech.com/AV/Converters/Video/USB-S-Video-and-Composite-Video-Capture-De...udio~SVID2USB2
Haugpauge USB-LIVE2
Both of these are usb capture devices. I am kind of stuck there as I need to capture onto laptops, long story, but a requirement for these configurations.

Computer: 3rd generation Intel I7 running Windows 7 or 8.1 (dual boot)

Neither capture device supports the resolution of the camera, so I am capturing at a resolution of 720 X 576 (highest PAL the cards support)
So the video looks ok, but speed seems to cause issues. It looks to me like an interlacing issue.
I have played with Aforge, and DirectShow. I prefer to code in C#, but can make due with C++ if need be.

So Aforge is perfect for my needs using the New Frame event handler is just what I want, and perfect. However the video is interlaced, and I don't know how to fix that on Aforge, or to show a deinterlace into its graph, or whatever.

I have used directshow to manually create the graphs (paaaainful).
DirectShow
Crossbar device -> Capture Device > (call render to create an output pin) -> Deinterlace.ax (DScaler Deinterlacer) -> Sample Grabber > NullRenderer

This kind of works, but sample grabber seems to intermittently fault on me.

Any ideas ? I don't want to be a video expert, I simply want the PAL camera input to create a Bitmap file (deinterlaced) that I can then process ???
I have googled aforge like crazy and the only comment I find from the author about deinterlacing says RTFM, which I have done over and over.

I know I shouldn't have to fork out 3k for an NTSC camera just to make this work, but about ready to at this point.

Any ideas ? Also my app is 64bit, I am forcing it to be 32bit because the DScaler is only 32bit.
Thanks so much for any advice

[jagabo]

Each frame of interlaced video (PAL or NTSC) contains two separate half-pictures. The two half pictures are interleaved line-by-line. So one is image is contained in the even numbered scan lines, the other in the odd numbered scan lines. Deinterlacing methods vary from a simple bob (separate the two fields, fill in the missing lines with data interpolated from the lines above and below) to complex, motion interpolated techniques that fill in using data from other fields when possible.

There are a lot open source programs with sample deinterlacing code.