Color Space relationships - HD(V) to SD DVD

[Stephen L. Noe]

Hi,

After alot of research I've come to the revelation that the ITU 709 colorspace needs correcting to ITU 601 for every DVD created. Paolo's 'true color' scene file baffled me for a long time. Why is the DSC chart and the waveform reflecting the full gamut and yet everything I produced using that scene file is way too saturated? The solution I see is that the 709 color space (YPbPr) will require a color correction to bring back to a linear relationship as in the 601 color space (YCbCr).

I believe the gamma should be corrected by a factor of 2 or 2.8 and then the gain for the individual colors should be adjusted to reign in saturation and then the gamma should be adjusted once more for the final output. I've been doing this instinctively and I think if anyone else can try this you'll find the blacks come in line as well as the reds. Try doing a color separation and you'll find the GBY is beautiful but the R is way oversaturated and the blacks can suffer.

Anyone else have experience converting color spaces? What is your rule of thumb?

[John M Burkhart]

Thanks for this Stephen,

I've been a bit stymied by the conversion from HDV to SD DVD, myself. I did notice that in a straight conversion the colors just didn't look right.

I've been kind of making do with down converting a HD clip of color bars, color correcting them on the vectorscope in SD, and then applying that to all my clips as a base correction. Then making individual color corrections on each clip.

Just a workaround for now, but can't wait until I can deliver on HD and bypass the whole process.

JB

[Steve Mullen]

Originally Posted by Stephen L. Noe

After alot of research I've come to the revelation that the ITU 709 colorspace needs correcting to ITU 601 for every DVD created. Paolo's 'true color' scene file baffled me for a long time. Why is the DSC chart and the waveform reflecting the full gamut and yet everything I produced using that scene file is way too saturated?

You are quite correct -- the wonderful side of HD (709) is that you can have very bright, saturated color. Color we could never have in NTSC or PAL.

Of course, our DVD systems are still NTSC and PAL. So you need to do a 709 to 601 conversion -- which is likely more than simply lowering saturation.

My guess at a solution -- record a Macbeth color chart on both DV and HDV. Then iterate color correction on the HD version until you get something that matches the DV version.

Now, SAVE that correction -- and apply to a copy of an HD timeline before exporting to DVD.

[John Jay]

Revelation or Noe; see here :)

http://www.dvinfo.net/conf/showthread.php?t=65935

[Chad Terpstra]

Does anyone know if there is such a correction (709 to 601) in Final Cut? I was unaware of the two color spaces until recently when a down-converted clip (HD sequence nested in an SD sequence) showed extremely different results in the vectorscope (much more saturated if nothing else). I thought it might be a fluke, but now that I know there's something I need to do, I'd be interested in knowing exactly what.

[Keith Winstein]

Originally Posted by Chad Terpstra

Does anyone know if there is such a correction (709 to 601) in Final Cut?

Hi Chad,

Yes, check out volume III, page 552 of the Final Cut Pro 5 User Manual:

"High definition video formats such as DVCPRO HD often use colors defined by ITU-R BT.709, or Rec-709. Standard definition formats use a color space defined by ITU-R BT.601, or Rec-601 for short.

"Final Cut Pro automatically handles both color standards, and accurately converts from one standard to another when you mix formats. For example, if you edit a standard definition DV clip into a sequence that uses a DVCPRO HD preset, the standard definition media is properly converted from Rec-601 to Rec-709. High definition video used in a standard definition sequence is also converted properly from Rec-709 to Rec-601."

[John Mitchell]

I can tell you that technically both the colour and luma are compressed in the 601 colourspace. I've never read the 709 spec but I assume it displays the full gamut of RGB colours (although it stores them as YUV for efficiency)
I'd appreciate it if someone who does know the 709 spec well can confirm this? Anyway based on that.....
If we assume 255 units as the total luma in RGB, you lose about 16 either end in 601. That's about 6.27% either end. So you need to boost the blacks by 6.27%, crush the whites by 6.27% and reduce colour saturation by about the same amount (colour is limited to roughly 238 on the RGB255 scale in 601 - handy for Photoshop). Many encoders like ProCoder have a built in converter expand to 709 from 601 and vica versa. I would rather do this at the encoder stage than individually colour correct timelines, but you have to have faith in your encoder - and the problem is computer displays are different again - so you can't "judge" the final result from what you see on your computer screen. Always check your final product on a broadcast monitor.
Gamma should definitely by lifted for computer displays if they are your intended output as they are "darker" or more contrasty than PAL displays (and I assume NTSC).

[Stephen L. Noe]

John,

Your post eludes to exactly what I'm seeing. I cut the video on an HD timeline and then if I produce a DVD the colors are linear"ized" to the 601 color spec, however I'm editing on a 709 timeline and previewing on an HD monitor via VGA second head. In other words the timeline and output from the timeline are being cut and previewed in the 709 space.

Liquid brings the colors into 601 either using "soft clipping" (which works more like clamping and forces all the colors into the 601 gamut by adjusting colors in relationship to one another) or just plain clipping (which brings everything under legal 601 by clipping off anything outside of it's gamut).

This is something to be aware of if you're editing completely in HD and then delivering in SD. The colors may become oversaturated on SD-DVD with Paolo's True Color scene file which is at max saturation for the 709 space to begin with. This is why I had trouble with skin tones using the great file Paolo came up with.