handling 16-255 camcorder video

[2Bdecided]

In the light of this thread...

http://forum.doom9.org/showthread.php?t=164981

...I'm wondering what method(s) of handling typical camcorder levels (16-255 luma, goodness knows what chroma) actually give something with matches the "original" colours, when viewing the YUV, decoded to 0-255 RGB, on a PC display?

Maybe just using levels(0,1.0,255,2,235,coring=false) to get in-range YUV
is closest? It works well enough in practice, but I'm wondering now if it's "correct"?

Does anyone know what camcorders do internally to generate this range? Are they creating 16-255 RGB and then generating the 16-255 Y directly from that? Or are they creating 0-255 or 16-235 RGB, and then using strange scaling to generate the 16-255 Y? Does it matter?

It may be a fair assumption that, whatever the camcorder is doing internally to create this strange range, it looks "right" when displayed on a TV that handles this range OK (especially old CRTs - the default display technology when this trend started) - and the closest you can get on a PC is to make it all a little darker.

Cheers,
David.

[Mounir]

Can we see the footage , a sample is welcomed

[2Bdecided]

It's an abstract question.

The three different camcorders I've owned, from different manufacturers, bought in different decades, using different formats, have all done this. AFAICT it's ubiquitous with consumer camcorders.

...but I'll try to dig a sample or two out. Plenty of samples to choose from, but most have people in them, who probably don't want to appear on doom9 (or anywhere else!).

Cheers,
David.

[poisondeathray]

Does anyone know what camcorders do internally to generate this range? Are they creating 16-255 RGB and then generating the 16-255 Y directly from that? Or are they creating 0-255 or 16-235 RGB, and then using strange scaling to generate the 16-255 Y? Does it matter?

I remember reading somewhere from (maybe from a engineer??) that DSP varies immensely between cameras and settings (e.g. if you select a certain color profile, or even if it doesn't have profiles).

Essentially a custom LUT is applied and none of them - not even run of the mill consumer cameras - are standard ITU BT 709 or 601, that's why superbrights are often preserved

I don't think it really matters ,unless you want to apply a LUT in camera (certain profiles)

[2Bdecided]

I usually just do whatever looks OK.

But I've just triedlevels(0,1.0,255,2,235,coring=false)
...followed by this...
http://forum.doom9.org/showthread.php?p=1403732#post1403732
...and it's amazing how many saturated colours captured by the camera are still out of range. This is from a bog standard Sony DV camcorder.

So maybe nothing "correct" is going to cut it. As you say, the cameras just do what they want. Some older cameras don't seem to care about legal RGB at all! Fine on old TVs, but hopeless on PCs.

FWIW some flat screen TVs still seem to work in Studio (16-235) RGB, giving quite a bit of headroom.

Cheers,
David.

[TheSkiller]

I know this behaviour very well from my own camcorders. To me it seems like this: the camcorder adjusts it's exposure and anything that exceeds it's aimed maximum is simply not cut off, giving this super white headroom that does indeed carry usable picture information.

I always make use of this as I found it to be an useful extension of the dynamic range that was captured.

If I plan to edit the footage in a NLE I convert everything to RGB anyway and save it to a lossless Ut Video Codec AVI. So I just do something like this:


Levels(16,1,255, 0,255, coring=false) #I actually use SmoothLevels
ConvertToRGB(matrix="pc.601") #or use dither_convert_YUV_to_RGB

Of course this assumes that the colors are 16-255 as well (don't know if they are, but I suppose) but this always worked very well for me.


Edit: I also found that CRTs do display the super white properly if you play back the raw tape via the camcorder's outputs.

[um3k]

Even my DLP rear projection displays superwhites correctly, oddly enough.