4:2:2 cameras?
 

Which camera's sample at 4:2:2? I don't usually see the sampling rates advertised but doesn't this add allot to the quality of the image? I plan to do some composite work, would this sampling give me better mattes?

Bryan

At normal viewing distances the difference is very small. Adam Wilt's DV FAQ might have something on this.

I don't think the added color information would be useful for compositing unless making your video bigger. If you're doing chroma keys then 4:2:2 would be useful.

You won't find any 4:2:2 sampling in anything less than Beta SP/Digibeta. I think DVCPro50 may also use 4:2:2.

4:2:2 will give you better matts, better rotoscoping , better keying, color correction , titles . BETTER IMAGE too.( period)

I agree with Don on the value of 4:2:2 in production and post. However, you will not find cameras with 4:2:2 in the prosumer price class.

Actually Hi-8 is 4:2:2 IIRC as is S-VHS and maybe even VHS and 8mm.

My Hi-8 footage chromakeys quite nicely.

Any of the 50 Mhz digtal formats are 4:2:2 (NTSC) including DVCpro50 and D-9.

Editing systems that immediately convert the DV image to 4:2:2 do a fairly good job of chromakeying, image processing, titles, etc. Mine feeds the 4:2:2 video out the analog ports and doesn't revert to 4:1:1 until it has to output via the firewire port.

Digital term - analogue media
 

Mike,

all of Hi-8, VHS, S-VHS (and 8mm film) are analogue formats, so 4:2:2 (and similar) terms don't apply.

An analogue source is just as good as its bandwidth. Yes, they may have as much chroma bandwidth as luma (so are capable of giving higher quality capture), or they may have less.

When you capture (digitise) an analogue source like this, the capture route you choose will produce 4:1:1, or 4:2:0, or 4:2:2, or even 4:4:4. Standard DV cards will give 4:1:1 or 4:2:0 depending on whether you are NTSC or PAL. Digisuite gives 4:2:2.

Of course, the quality of the resulting digitised footage will depend on the bandwidth of the analogue source. You could capture VHS to 4:4:4 if you wished, but it would be a waste of hard disk space.

Having said that, it is true that analogue footage often keys very well, if captured correctly. This is because it has smoothly changing chroma, and not the overly blocky chroma of 4:1:1/4:2:0. As I've mentioned elsewhere, I personally use an old 1980's Ikegami analogue colour camera to capture direct to PC (via a DPS Perception card, with no tape used in the camera) for keying - it works beautifully.

Of course you could upsample your 4:1:1/4:2:0 to 4:2:2 or 4:4:4 before keying - it will give a better key than the 4:1:1/4:2:0, but will be blurred a bit compared to 4:2:0 or 4:4:4 original.

Regards,

Julian

Ok, as long as we're on the subject, someone please help me out because I am confused.

As far as I know, 4:1:1, 4:2:0, etc refers to the decimation of the color samples at capture time INSIDE the cammera...that is, the 3-CCD's produce complete color, but for a 4:1:1 system, full luma is recorded but color is only recorded every 3 pixels.

Now, my question is, when one captures the DV stream into a computer, why would it matter how the camera decimated the color? After all, all the computer gets is a series of pixels for each frame, and once it is in the computer in 8-bit color, what is the difference? I guess somewhere along the capture process, either the computer or the camea approximates the missing color data, but once it's in an 8-bit or 16-bit uncompressed format for editing, what difference does it make if the camera originally threw away some color samples as oppossed to others?

I expected to find the solution when i bought my PAL XL1se, but it turns out that i still don't understand it...I recorded footage in PAL, resized it and adjusted frame rate to NTSC and the DVD's look perfectly fine.

Thanks,
Juan

Julian,

Just got back to this subject. I do believe sampling rates do apply in analog recording. Bandwidth limits do apply to the recording mechanism and one of the way to get more resolution was to drop the amount of information storage required of the medium.

I seem to recall that even BetaSP is only 4:2:2 in the NTSC flavor. But I could certainly be wrong but it was an issue 5-6 years ago and highly discussed then.

If the sampling is missing data it is more likely to show posterization, especially in areas of solid color. Look at a histogram in PS when you adjust levels. Data is lost because of the need to correct poorly sampled data. Rounding errors are also more likely.