Using PAL-DV superwhites & superblacks

[freezer]

Hi!

I have one question:

I am using my own footage shot with the Canon XM-2 (PAL). After capturing I do some filtering with AVS (reinterpolate420, deinterlace, denoise). When I look at the levels filter in VDub, I can see that I have some white clippings. So I also use
ColorYUV(levels="PC->TV")

but this seems to push the blacks too much (it works well for the whites). I must level the blacks down by 8 to get them back.
Are there any methods to improve my workflow?
I did a search, but found nothing helpful...

Thanks in advance!

[juhok]

I don't know about XM-2 but the 3 Sony consumer DV cams I've used don't really have "TV" levels. They produce something like 16-255~. So simple TV->PC or PC->TV conversion is not doing much good.

I'd try to analyze parts of video with pure black and white and determine the "real" levels and go from there. It also might be your DV codec decoding it improperly etc.

These are just my observations, take with grain of salt.

[freezer]

The XM-2 definately uses levels above 16 and 235. I am using the standard MS DV codec, where I have no possibility to change level settings.

Shall I provide some samples to play with?

EDIT:

Here is an example footage (http://www.loom.at/~loomdata/private/XM2_Example_footage.avi). (4 secs, 16,6 MB)
PAL 720x576, interlaced, shutter 1/50sec

[juhok]

I copied some frames to PhotoShop and it looks like white is maxed out @ 255 and highlights are lost with every step "backwards". Black is almost at 0, the frames I tried could use 8->0, so I would do something like:

Levels(8,1,255,0,255, coring=false)

But with some other frames black could be closer to 0 and I would go as far as to say that XM-2 in this case gives full "PC" range and doesn't necessarily need any conversion whatsoever. Pretty strange behavior from a professional camera IMHO.

I'm using Cedocida DV Codec 0.17 (YV12, MPEG2 Interlaced Chroma)

I'd really like to have second opinion on this. :p

[um3k]

Using cedocida to decode, this brings the whites into range:

ColorYUV(-20)

It can't do anything for the parts that clipped completely, but it will bring back whatever detail exists. It has no noticeable effect on the blacks.

Also, you don't need to use reinterpolate411* on PAL footage. It's meant for NTSC DV, which has the color encoded with 4:1:1 subsampling. Avisynth doesn't support 4:1:1 (yet), so the video must be converted to 4:2:2. PAL DV is 4:2:0, a.k.a. YV12, which is natively supported in avisynth. Cedocida codec feeds the video to Avisynth in this format, which allows maximum quality to be preserved.

*I assume you're talking about reinterpolate411 since "reinterpolate422" doesn't seem to exist.

[Wilbert]

I guess he means reinterpolate420.

[um3k]

Ah. Never mind about that part then...

[freezer]

Using cedocida to decode, this brings the whites into range:

ColorYUV(-20)

It can't do anything for the parts that clipped completely, but it will bring back whatever detail exists. It has no noticeable effect on the blacks.

Thanks for the hint - it improved things a lot!