hdlink & avisynth input
 

I've been testing out the trial of NEO and wasn't very happy with the deinterlacer built into HDLink. So I wanted to use aviSynth so I could have better quality and was please to find out that HDLink would indeed accept avisynth file.

I searched on the internet and I didn't see anyone else ever doing this so I thought I would post that it indeed could be done.

One question, I had to convert the colorspace to RGB or YUY2 from the default YV12. Should I use one over the other?

If your source YUV, stay in YUV, as converting to RGB requires the correct 601 or 709 matrix math (most do this wrong.) Therefore convert YV12 to YUY2 (which is faster anyway.) Of course if you source is RGB then leave it that way.

P.S. Please tell us more about your AVISynth workflow.

Dont take this negatively, but ... why would you feed an avisnth script into hdlink? If the script is capable of opening the HDV mpeg2, then you can process and save it to a cineform avi using Cineform's vfw codec via Virtualdub which gives you a visual GUI etc.

There is nothing really too it. I run the m2t video through DGIndex.
http://neuron2.net/dgmpgdec/dgmpgdec.html

Write up a little avisynth script http://avisynth.org/mediawiki/Main_Page

My avisynth scripts looks like:
v = MPEG2Source("07-10-27-ace-mgmb.d2v", idct=3)
a = MPASource("07-10-27-ace-mgmb.mp2", normalize=false)
AudioDub(v, a)
Load_Stdcall_Plugin("C:\Program Files\AviSynth 2.5\plugins\yadif.dll")
Yadif(order=1) #deinterlacer, one of may available
LanczosResize(1920, 1080) #resize if you want
ConvertToYUY2()

And open the avisynth script in HDLink and convert to cineform.

The only reason you would go through this is if you wanted to preprocess the video. In this case I wanted high quality deinterlace and resize.

You are right I actually used VirtualDub first. But I just wanted to see whether HDLink would work. And when it did I thought I would let other people know.

Some people may be more comfortable using HDLink.

Plus to be honest I don't know how to batch process videos in VirtualDub, something HDLink does easily.

Eric,

so that you aren't thinking there is anything work with the HDLink resizer, is also Lanczos. The deinterlacer you did like is higher quality than most for the amount of resolution it can extract, but it does introduce more motion blur as it is using both fields for creating the new frame. It is a linear light algorithm so the motion blur it does introduce is much more natural looking than other two field deinterlacers, although it does prefer that the source shutter be at the frame frame, 1/60th for 60i acquistion, 1/50th for 50i etc (fortunately this is the typical camera default, highly recommend for 60i->24p.) We are thinking of adding a single field deinterlacer, for those who prefer that look, and those using higher shutter speeds (for the stroby look or they don't have ND filters,) and the often requested 60i to 60p for later slow motion work. What type if deinterlacer are you looking for?

>>Plus to be honest I don't know how to batch process videos in VirtualDub, something HDLink does easily.

Ah - good point!

I would find the 60i to 60p very useful and a good 60i to 30p that can deal with motion, panning and motion in the scene.

I shoot dog sports, and I'm sure the HDLink deinterlacer is probably more geared toward your typical static film type shot.

As you will see in the pictures of my footage the built in deinterlacer does not handle it very well. Which I understand you can't be everything to everyone.

Set your shutter to your frame rate and the double image will go away on most cameras. Then you get the benefits of normal look motion blur and more vertical resolution. If the top field is exposed for 1/60th and the bottom field the next 1/60th, together they produce a smooth 1/30th shutter (with the right deinterlacer.) It is true our deinterlacer is targetted for indie filmmakers, if they have to shoot 60i, they will normally shoot locked down to this shutter speed / exposure time.

For 60i to 60p, you might take a look at this:

http://www.guthspot.se/video/#deinterlacesmooth

I shot some footage today using 1/60 shutter and as you suggested HDLink deinterlacer worked extremely well with it.

Am I correct to say that the Cineform deinterlacer basically blends together the two fields. And in effect this would be as if you shot 30p at 1/30 sec shutter?

That is correct, it does it in a way to extract a lot of vertical resolution. One additional note, most cameras can't do this, and the 1/30th shutter mode is only one field worth of resolution, so it is not completely equivalent.