8-bit color RGB versus 10-bit color 4:2:2

[fouronthefloor]

Hey guys this is my very first post on this forum. I've been lurking on here for a while, and I've learned a lot by reading old threads, guides, etc.

I have some very important Hi8 footage that I need to capture urgently. The tapes are 15 years old now, and I want to archive the analog footage into video files with the BEST QUALITY possible. $ and file sizes is not an issue (provided I'm not throwing it away for no gain). The question that's confusing me right now is the choice of video format for archiving purposes.

After reading a lot of the threads here, I had my heart set on lossless Huffyuv set for RGB colorspace. I considered YUY2 colorspace, but I'm being VERY ANAL about quality preservation here. I don't want the captured video to have "chroma subsampling" distortion. Futhermore, I like to use Adobe premiere, and I read that internally, Premiere uses the RGB colorspace for all of its operations. I want to avoid colorspace conversions when I eventually get around to editing. I want to leave all lossy operations till the very end when I decide on the distribution format.

And then I started researching capture cards, and I stumbled upon Matrox's MXO2 Mini product. If I understand correctly, the product can capture to some uncompressed 10-bit color 4:2:2 chroma subsampled format.




Here's the question of this thread. Which format is better quality???


8-bit color depth/RGB colorspace

vs

10-bit color depth/4:2:2 (??? colorspace)

[jagabo]

You're stressing about this too much:

The 8-bit vs 10-bit ramp image is highly exaggerated. The top red gradient is really 8-bit color. The bottom gradient is between 4-bit and 5-bit color. There's so much noise in your Hi8 tapes that it doesn't really matter if you caputure as 8 or 10 bits. The lower order bits will just be noise. If you plan on doing a lot of filtering with software that supports 10 bit color you will get slightly better results (after filtering) with 10 bit caps.

4:2:2 YUV is closer to the true state of analog video (all capture devices capture YUV internally). But consumer tape formats have very little chroma bandwidth. So you're really looking at something like 4:0.4:0.4 in your Hi8 tapes. In other words 720x480 4:2:2 will capture all the color resolution of your tapes.

You have to be careful about YUV to RGB conversion. With standard definition video this is usually done with a rec.601 matrix that results in a contrast expansion (Y [16-235] to RGB [0-255]). If your 8-bit YUV caps (and remember the capture device is capturing YUV, even if it is set to save as RGB) contain any luma values less than 16 or greater than 235 they will be lost on conversion to 8-bit RGB. There will be no chance of bringing out details in those very dark and very bright areas.

So your best bet, if your hardware and software fully support it, is to use 10 bit YUV 4:2:2. But there will be hardly any difference between that and 8 bit YUV 4:2:2.

[fouronthefloor]

Thanks for much for your response jagabo. I don't understand everything here, due to my having limited knowledge of broadcast engineering. Please tell me if I got the gist correctly:


* For Hi8 tapes as source, a 10-bit capture won't offer significant additional image information over a 8-bit capture.
* However, investing in a 10-bit capture will make for better quality video restoration processing later.
* Hi8 tapes have a relatively low color resolution of ~4:0.4:0.4. Capturing at 4:2:2 should pick up all of the color information from the tapes.
* The YUV-->RGB colorspace conversion is a lossy transform. Capture hardware does it's capture and processing in YUV internally. Saving the capture on the fly to RGB will actually force the lossy transform?



PS to anybody: Is there a great broadcast engineering textbook with all the gory math details out there? I'd like to study up on this stuff on my free time for a more formal introduction on the subject.

[Gavino]

Is there a great broadcast engineering textbook with all the gory math details out there? I'd like to study up on this stuff on my free time for a more formal introduction on the subject.

Digital video and HDTV: algorithms and interfaces, by Charles A. Poynton

[jagabo]

Please tell me if I got the gist correctly:

Yes to all.

Poynton's web site: http://www.poynton.com/Poynton-video-eng.html