4.2.2. versus 4.2.0

[Kevin Shaw]

Originally Posted by Thomas Smet

If you zoom in and look at the edge of any footage you will notice how the edges have gradation and are not pixel sharp.

That would certainly be true for any camera which doesn't have a 1:1 mapping from the sensor to the recorded and output video formats. The EX1 solves this problem, except to the extent that interlacing affects the image. That's the point of my example: the recorded image from the EX1 potentially could be that sharp.

Maybe I'm missing something, but I don't see how video with blurry details would be easier to chroma-key than one with crisp details. Isn't part of the point of chroma-keying to identify the edges?

[Thomas Smet]

Originally Posted by Kevin Shaw

That would certainly be true for any camera which doesn't have a 1:1 mapping from the sensor to the recorded and output video formats. The EX1 solves this problem, except to the extent that interlacing affects the image. That's the point of my example: the recorded image from the EX1 potentially could be that sharp.

Maybe I'm missing something, but I don't see how video with blurry details would be easier to chroma-key than one with crisp details. Isn't part of the point of chroma-keying to identify the edges?

No the point of keying is to put something somewhere where they were not. The art of compositing is that your final result should look exactly how it would have looked if you shot that person in that environment. The more crisp an edge the more you will notice low chroma resolution. It also depends on the quality of the keyer software. Some simple keys yes they only works in single colors so a harder edges will give you better results but these are not great keyers. They leave the edge looking fake with no smooth edges or even motion blur for that fact. Real video and film has motion blur and a good keyer should be able to deal with that. I hear some people talk about shooting with a high shutter speed so there is very little motion blur but that isn't very realistic. If you shot your actor in that environment would you have shot it that way? Sure the edges of the key may be clean with cheap tools but it is fake looking.

With the proper keying tools the softer image will look more natural unless the camera really was that sharp. No the EX1 will not be that sharp either. Even the F950 isn't pixel perfect 1:1 sharp. Video has to have low pass filters so the video doesn't have aliasing. Interlaced has to be filtered even more to reduce interlace flicker. Even photographs are not super crisp like computer graphics. The image I made is about as bad as you are going to get with reduced chroma resolution.

[Kevin Shaw]

Originally Posted by Thomas Smet

The image I made is about as bad as you are going to get with reduced chroma resolution.

In which case I don't see a problem for the EX1 when used in progressive recording mode, as the third image in your example looked fine without magnification - only the interlaced example really stood out as having issues. So running the EX1 in progressive mode and editing the resulting footage in a 4:2:2 color space shouuld work pretty well for normal viewing circumstances.

[Mike Schell]

Originally Posted by Kevin Shaw

Mike: since the Extreme III cards are supposed to support sustained write speeds of 20 MB/sec or 160 Mbps, can you go beyond 100 Mbps reliably with your device on those cards?

Hi Kevin-
We won't know until we do a lot more tests. The 20 MB/sec assumes sequential writes, which is possible 95% of the time using the MXF OP-1A format. We do have some non-sequential writes that must be performed even with this fomat (header info). Also, you need to allow some extra margin for audio data, header info as well as updating FAT tables and opening / closing files. So, we rather quote a more conservative figure until we can do more exhaustive tests.

The good news is that Compact Flash write / read speed does not degrade as the card fills up, unlike hard-disk drives.

Mike Schell

[Kevin Shaw]

Thanks Mike. Do you have a standard recording option with a bit rate somewhere between 100 Mbps and 160 Mbps, or is 100 the main conservative choice?

[Thomas Smet]

Originally Posted by Kevin Shaw

In which case I don't see a problem for the EX1 when used in progressive recording mode, as the third image in your example looked fine without magnification - only the interlaced example really stood out as having issues. So running the EX1 in progressive mode and editing the resulting footage in a 4:2:2 color space shouuld work pretty well for normal viewing circumstances.

I agree 100%. That is why I posted the images. progressive 4:2:0 is a very clean format for keying. Sure it isn't perfect but it is pretty darn good actually. I forgot what editing system you use but if you have a way to soften or up sample the chroma then it will be even better yet. By the simple chroma soften filter I used the edges look just as clean for all the formats. Of course if you use a XDR you may not have to use a filter which means no darker edge and footage you can work with right out of the box.

One other thing to keep in mind about interlaced 4:2:0 is that it works out pretty good if you plan on keeping it interlaced. Each field gets played one at a time so the alternating samples do not show up when played back that way. That is why 60i 4:2:0 keyed as 60i 4:2:0 may look bad in the NLE but it should be fine once it is played back on a TV. Well it isn't perfect but much better then what we see here. The problem with interlaced 4:2:0 is when companies try to put progressive frames in a interlaced mpeg2 file such as the Canon HV20. So for those who want to shoot 60i as 60i the 4:2:0 should be fine although I personally would prefer 4:2:2 for 60i.

[John Mitchell]

Originally Posted by Michael Javorka

My situation is unique and maybe someone can enlighten me a bit. I come from a background as a 35mm film shooter, without having edited or managed the end use/workflow. I'm moving to HD and have learned a ton of new technical information already. But what is the difference between 4.2.2 and 4.2.0? I understand the soon to be released Sony XDCAM EX Cinealta will deliver 4.2.2 under a certain workflow.

I chiefly produce very high end stock footage, destined for my agent, Getty Images, who distributes and markets my material worldwide. It needs to be the highest quality possible to be intercut into movies, commercials, and documentaries.

Michael - I would be strongly advising against the EX as a replacement for 35mm film. There is no doubt the EX is a revolutionary handycam, but at best it still has many of the limitations of a handycam - not the least 4:2:0 or 4:2:2, but how about a fixed lens? half inch DOF as opposed to 35mm? CA that you would only find on cheap 35mm lens, etc. You could consider it as an option where lugging around a 35mm kit was impossible.

I really think RED is a much better match for what you are doing - it's more expensive, but you gets what you pay for - and I don't know how much shooting you do but it would surely pay for itself fairly quickly. It a full sensor, with a 4:4:4 post workflow pretty much designed and working.

If you are unsure I would rent first, decide later.

[Alexander Ibrahim]

Originally Posted by Thomas Smet

It should also be pointed out that HDCAM tape uses 3:1:1 color and only samples 1440x1080 pixels. This give you chroma channels that are 480x1080 pixels in size and yet a lot of people pull great keys from this tape format and the tape format is regarded as the industry standard.

I've said elsewhere that the absolute number of chroma samples isn't the determining factor, but the ratio of chroma samples to luma samples. Thus the choice of terminology being expressed as a ratio.

So, by this measure, 3:1:1 should give a better key than 4:1:1 or 4:2:0. 4:2:2 is better than 3:1:1. 3:2:2 (which to my knowledge has never been used) would fit between 4:2:2 and 4:4:4. 4:4:4 and 3:3:3 (another format that is, to my knowledge, mythical) would be exactly equivalent.

On second thought, calling 3:3:3 mythical is off the mark. We use it often- we just call it 4:4:4.

So if 3:1:1 was good enough for ILM and Episode 2 I'm sure the rest of us can make great use of other decimated chroma formats.

But 3:1:1 wasn't good enough for Star Wars.

They recorded 4:4:4 off the head to outboard decks. (D5-HD decks I think.) That said, I bet there were some shots where they said "heck with it, pull the cables." Then HDCAM would rear its head.

For Star Wars 3, Sony delivered HDCAM SR. I think the dual link HD-SDI spec was developed to fill the technical needs for Episode 3.

[Alexander Ibrahim]

Originally Posted by Thomas Smet

I also wasn't talking about keying ability but the fact that progressive 4:2:0 is not garbage. In it's raw form 4:2:2 is better but with filtering the results are going to be much closer.

I see we are agreeing contentiously.

4:2:0 isn't garbage.

There are a lot of circumstances where 4:2:0 will work very well. Even the interlaced version can work well for some productions.

In the end what I am deeply concerned about is for most users an edge case. (pun sadly intended.) Most stuff that people shoot and need keyed will be far simpler.

4:2:0 is inferior color sampling to 4:2:2 in all cases. In many productions it won't matter. 4:2:0 will produce exemplary keys.

When it does matter, the difference will be pronounced.

Because I like to throw wrenches into the works, I will add that 10bit vs. 8bit is a more important consideration for a lot of post workflows.

I don't know why you say my method takes a lot of time. It doesn't really take that much at all. A filter in FCP takes a second to drop in. In shake it just takes a few quick nodes which you can then save to reuse for the future.

It isn't building the tree that I think takes time. Tweaking all the parameters of all those nodes for best effect is what I think will suck up artist time.

On the render side- the built in nodes are much more efficient than scripts we build using the interface. The fewer nodes we use the faster our renders will be.

So, its a double whammy.

Now that we've beaten this horse to death, I'd just like to say that proper planning, the right tools and smart craftsmanship of the image can do way more to make the image work for audiences than just about any technical detail.

[Alexander Ibrahim]

Originally Posted by David Heath

Hmmm. I think most of that really is dreaming....... :-)

Whereas 4:4:4 seems perfectly feasible within proven technology - just 50% more bits than 4:2:2, keeping all else equal.

Desirable though much of your wishlist may be, I'd expect a general move to wavelet compression (as used by Red and JPEG2000) to be the next big step.

I agree that wavelet compression is the next big thing.

I disagree with the long term outlook.

The question isn't whether or not most of my wishlist will come to pass, but rather when.

Storage is getting cheaper, bigger and faster very quickly. Let's have a trip down memory lane. (once again I intended that sad pun.)

http://www.pcworld.com/article/id,127105/article.html

http://upload.wikimedia.org/wikipedi..._over_time.png

1956 IBM introduces the hard drive, with a 5MB drive the size of two refrigerators.

1980 IBM brings us the first 1GB hard drive... at a weight of 550 pounds and selling for a paltry $40000 USD.

1991 IBM Brings us the first 1GB 3.5" drive.

2001 Western Digital's 100GB drive sold for $449

2007 HITACHI, the former IBM Hard Disk division, introduces a 1 TB drive. You can buy it now for about $280

http://www.bestbuy.com/site/olspage....ci_sku=8295588

I predict, by extrapolating these curves, that we'll have 10 TB drives by 2011. By 2021 we'll have 10 Petabyte drives. That's 10000 TB- the same 1000 time improvement we got from 1991 to 2007. Basically I am betting on Moore's Law and its analogues in storage tech.

The exact numbers don't matter much. I'm going to simply assume that we eventually get a 1 petabyte drive.

A little fast envelope math. Uncompressed 10 bit 4:4:4 HD SDI dual link is about 3 Gbps. 4k is about 4 times the pixel resolution of HD and thus four times the data. (Yes I know its a bit larger but 2K and HD 1080p are pretty close in a 16:9 format.) 8k is about 4 times the resolution of 4K so...

Uncompressed 8K 16:9 video at 4:4:4 10 bit should be about 48Gbps. My 16 bit log version should run about 77Gbps, I'll round that to 10GB/s. That's 1 min 40 seconds minutes of storage per TB. I am going to round that down to 1.5 minutes per TB to handle various overhead in the file system etc.

That's 1500 minutes per petabyte. 25 hours of uncompressed 8K 4:4:4 16 bit footage.

Now to accomplish all this we need drives that write a sustained minimum of 10GB/s. That's about a 600 fold increase in speed from today's. (A lot of single SATA drives transfer about 1GB a minute. There are faster drives, but I'm using that number for the basis of my math.)

So when we first get Petabyte storage around 2015, in an array of say 7 150 TB drives, we'll be dependent on RAID-0 to get the speed we need.

By 2021 I think we'll be able to stretch a smaller 2-4 disk RAID 0 volume to do the job.

By 2030 A single drive will handle 1PB at uncompressed 8k 16 bit 4:4:4 with hours of storage.

We live in a weird place technologically. We live where compression is required to get basic video work done for most people without exorbitant budgets. We live where storage space is an issue.

This will be a historical anomaly, most of history will be recorded using standard uncompressed video formats on standard and highly reliable media.

Today, we struggle without those future standards, and we struggle to invent them.

Apple however can be counted on to provide us with a low end laptop and a high end- with no middle of the range. To think otherwise I maintain would be madness. (Not that I would mind a bit of madness... in case anyone at Cupertino is reading this.)

[David Parks]

That's very informative. So maybe NASA will be shooting with Red 8k digital video from the moon by then. Ultra Hi rez images of a gray and tan surface.

Cheers.

[Kevin Shaw]

Originally Posted by Alexander Ibrahim

By 2030 A single drive will handle 1PB at uncompressed 8k 16 bit 4:4:4 with hours of storage.

And somewhere between now and then most movie theaters will probably switch to digital projection using some lesser form of compression, meaning there will be little need for that level of recording quality. And more people will be watching their movies on home TVs or portable video players requiring even less quality, although it never hurts to have high-quality source.

But as far as the EX1 is concerned, the native 4:2:0 will be fine for many of us and the 4:2:2 option with the Convergent XDR recorder should be awesome. If you really want 4:2:2 recording either buy an HD-SDI recorder or a more expensive camera, or settle for marginal resolution with the HVX200.

[Alexander Ibrahim]

Originally Posted by David Parks

That's very informative. So maybe NASA will be shooting with Red 8k digital video from the moon by then. Ultra Hi rez images of a gray and tan surface.

Cheers.

OK, I get it- yanking my chain. I'm laughing. :) I get that I'm off in la la land when I do that, but I find it amusing- I hope you do too.

Space scientists would love nothing more than to see these sorts of images projected so they can really stand on them. Of course... they may as well be black and white for lunar images, but on other bodies full color would be useful.

Originally Posted by Kevin Shaw

And somewhere between now and then most movie theaters will probably switch to digital projection using some lesser form of compression, meaning there will be little need for that level of recording quality. And more people will be watching their movies on home TVs or portable video players requiring even less quality, although it never hurts to have high-quality source.

Theaters survive by providing a better experience than people have at home. I'd ask if that will remain the case in the future. I suppose it will because, ostensibly, theaters can throw more money at the problem of presentation fidelity.

I also think we are headed for high resolution displays. For computer use, as an example, 30" displays are the largest useful to most people. At that point you have to turn your head to see the entire image in desktop use. If you need more screen real estate you might as well use additional monitors.

The trend will be to increase on screen resolution- driving dots per inch towards resolutions like we see in the print world. I think Microsoft and Apple agree because they are building resolution independence into the OS.

Instead of a 320*480 iPhone I expect that I'll have 1920*1200 in roughly the same size screen, thus raising the screen resolution from 163ppi to a laser printer quality 620 ppi.

I question the value of that trend for motion picture applications, but I see that it will be incredibly useful as someone who reads a LOT on computers. (Just as I like reading higher resolution printed pages.)

Regardless of whether its useful to the viewer, that technology is coming- and we have to start thinking about it. Will we just upscale our current images... or will we all shoot 4k and higher resolutions?

I agree that recording quality should exceed presentation/delivery quality by as much as practical. In an era of such incredibly high, by present standards, resolution what will contribute to quality?

But as far as the EX1 is concerned, the native 4:2:0 will be fine for many of us and the 4:2:2 option with the Convergent XDR recorder should be awesome. If you really want 4:2:2 recording either buy an HD-SDI recorder or a more expensive camera, or settle for marginal resolution with the HVX200.

Nice way to lead this meandering fool back on topic. :)

I agree that 4:2:0 will be enough for most projects. I think it may well be enough for most scenes even in a film that will endure a great deal of post.

Unless you are shooting a visually stylized piece like 300. All that post contrast and crushing makes for evil contouring in 8 bit footage. (300 was shot in 35mm and scanned in a 16 bit log format- .dpx I think.)

Now that I've looked more into the Convergent Design Flash XDR I think it will be a great product. My problem with it is that it records 8 bit formats.

I think that for my technical film style work I'll have to choose a 10 bit solution- 10 bit matters more in DI than 4:2:2 I think. I'm kind of bummed because I was getting excited about the Flash XDR. I am considering one for a B camera, but I think I'm gonna go with the AJA ioHD product on A cam- even though my total solution will cost much more that way and be more of a hassle to set up on set.

[Jeremy Kromberg]

Check out the CineForm Software Showcase. They are talking about making a 10 bit mini DDR. Let them know what is important to you.

[Kevin Shaw]

Originally Posted by Alexander Ibrahim

Theaters survive by providing a better experience than people have at home. I'd ask if that will remain the case in the future.

I'd say it's the already case that in some ways you can get a better viewing experience at home, so theaters are mainly surviving because people like to get out of the house once in a while. With millions of households owning HDTVs and now starting to buy HD disc players, the quality of the home viewing experience is increasing while the experience of going to theaters is arguably declining with all the ads and previews and other distractions.

One good thing about the EX1 is that it records at close to the maximum quality most people are going to be able to see outside movie theaters, plus you could potentially archive the master footage on Blu-ray discs and be able to play those. This is a handy coincidence for anything besides big-screen distribution, and should make it a good camera for projects going directly to customers (both home and corporate).