A few questions about color bit depth of final output.
 

i was watching the webisodes of Sanctuary (due to the hotness of amanda tapping & emilie the new chic) and noticed that in the wide shot compositions, the color depth in the background CG renders were so low that you could literally see the lines between each hue of color. that got me thinking...

i know there are limitations to both SD DVD&HD DVD&Blu-Ray. the spex do not currently include the "deep color 36 or 48-bit color depth". even if the spex included it, the current players can't output it. thus, my discussion will 100% be limited to PC output to be PLAYED back on the PC.

1. does it take more processing power to play higher color depth video?

2. i know CG can be rendered and outputted as whatever depth you want due to the digital nature, but hdv footage can't because when you capture live action, you're automatically limited to the colorspace medium you're capturing to, unless you can do solid state or direct to disk, which means you can do 4:4:4. thus, the question remains, how can you achieve higher color depth output of the final render if what you're capturing is low color depths?

3. let's say i make 1 master of highest color depth. i'm already converting the lower color depth footage to the higher color depths once. if i render FROM that higher color depths source/master, will i suffer generational loss of color since i'm going from low to high then back down to low?

PS when i get back home, i'll try to screencap what i'm referring to. perhaps my lingo is incorrect =P. i'm a n00b when it comes to color depth.

Yi Fong Yu,

Absolutely. If we assume uncompressed video, the I/O bandwidth requirements scale linearly with the color depth IIRC. Assuming compressed video, the I/O requirements become less of a factor, but the CPU requirements will increase as resolution and color depth increase.

Well, if you actually shoot HDV, it's in YV12 4:2:0 MPEG-2, no ifs ands or buts. If you shoot direct to disk off the HD-SDI spigot from the XL-H1 for example, you get 4:2:2 uncompressed, which is MUCH better. True 4:4:4 (RGB) can only come from very high end products like the F-950 IIRC.

The thing is, if you're doing any compositing, filtering etc, you need to maintain as much precision as possible. The thing to remember is that these filters work internally in high resolution colorspaces (usually RGB, or 4:4:4 YUV), and you don't want to throw away the extra precision generated by their calculations. So that's why you would render the output of the composite to RGB.

Of course, you're going to throw away all that info when you go down to 4:2:0 YV12 for MPEG encoding for DVD (even HD-DVD and Blue-Ray only support this colorspace). If you're talking about PC only, you will probably only find lossless codecs that can handle 4:4:4, RGB, or even 4:2:2. All others will probably only accept 4:2:0 (YV12) or will internally convert using god knows what black magic ;)

Technically, yes. Subjectively, probably not much. The only reason you would do this is to apply filters, as I stated above. Remember, garbage in garbage out. There's no point in upsampling footage if you're not applying any effects to it.

DV's horrible chroma resolution 4:2:0 for PAL and 4:1:1 can benefit from chroma upsampling to 4:2:2, or it will suffer from stairstepping and blocking that's especially visible in a situation like a red ball bouncing in front of a black wall. Most DV codecs will handle this when they decode (since they will usually output 4:2:2 or RGB), but some don't.

The artifacts you are seeing may be due to the compression... one compression technique is quantization / lowering the bit depth. (Though this is done on the DCT components, which looks different than lowering the bit depth on every pixel. The bit depth is lowered more for high frequency components than other components.)