hv20: bright colours outlined in pixel distortion

[Jay Cowley]

I'm finding that with my HV20, bright colours, especially red, show large amounts of little squares and weird pixels in video, when you see them in beside contrasting colours.

see attachment below.

has anyone else had this issue. do all these cameras do this, or just mine? anyone know what would cause this, how to get rid of it?

[Joseph H. Moore]

What you're noticing is a lack of color resolution (4:2:0) inherent in HDV compression, it is not unique to this camera.

If you were to look at the image "live" via the HDMI tap you would not see the same issue to the same degree, since the HDMI signal contains more color samples. (4:2:2) Even then, half the color resolution is being thrown away before being sent out.

Only a pure, 4:4:4 signal would have no chroma sub-sampling, and that is very rare, even among the most expensive formats.

[Wes Vasher]

Jay, you can 'clean' these color resolution or compression artifacts in software to a certain extent in your NLE or After Effects. Magic Bullet has a nice filter. Scroll down to 'Compression Correction'...

http://www.redgiantsoftware.com/mabued.html

[Robert Ducon]

Yup, it's the normal 4:2:0 compression - red stands out. All HDV cameras are like this, no worries.

I use G Film by Nattress to fix my colour:

http://www.nattress.com/Products/fil..._Sharp_Int.htm

[Jay Cowley]

really like some of this footage looks really bad, if there is a bright red option is is completed surrounded by these artifacts.

It just seems weird to me that all these big camera companies could release all these HDV Cameras, if they are completely flawed in that they cannot capture red. Is there something i'm missing here, that just doesn't sound right.

[Wes Vasher]

Originally Posted by Jay Cowley

really like some of this footage looks really bad, if there is a bright red option is is completed surrounded by these artifacts.

It just seems weird to me that all these big camera companies could release all these HDV Cameras, if they are completely flawed in that they cannot capture red. Is there something i'm missing here, that just doesn't sound right.

It's just the way it is. Watch bright reds on a DVD sometime, it's shocking. Bright red is something you try your best to avoid when shooting DV or HDV. It's not that big of a deal IMHO but there are work-arounds such as using HDMI and SDI capture as others in this thread have mentioned. Some higher end cameras record higher color resolution.

[Andrew Plumb]

Is it something you could work around, at least in bright-lighting situations, by adding a coloured filter to remove some red?

[Robert Ducon]

All DVDs are 4:2:0, so this 'completely flawed' system is standard world-wide.

Like we've said, clean the chroma with a 3rd party plugin. It's not rocket science, and since you're using a 35mm adapter, you prefer quality and are wiling to spend time/money for it.

Be sure to convert the HDV file to a different colour space and full 1920x1080 resolution - codecs: i.e. JPEG or ProRes 422 on the Mac). Otherwise, cleaning the chroma on HDV in an HDV timeline will keep the colour at 4:2:0 of course.

[Thomas Smet]

It's not just HDV or DVD. PAL DV also uses 4:2:0 color. All digital broadcasting via cable or dish is done the same way as well. 4:2:0 is just the way it is done in the video broadcast world. Yes high end formats can use 4:2:2 but all that footage eventually gets broadcast as 4:2:0 so they have to keep that in mind when they are shooting as well. Really 4:1:1 DV was much worse at it.

The reason why it stands out so much for you is because you are looking at a still frame on a computer. When you watch the tape again on a HDTV most of that should go away.

1080i cameras have a much harder time with 4:2:0 when the user plans on editing on a computer. 4:2:0 is done in a funky with interlaced video and when you dit you are dealing with whole frames and not fields like the video was meant to be displayed. This is in no way shape or form a flaw but just the way digital video works. Yet another reason why everything should have stuck with 720p and been done with it forever but lets not go there.

[Robert Ducon]

I should clarify, by flawed I meant tongue in cheek. ;)

HDV is great - go shoot!

[Mikko Lopponen]

HDV IS better than dv in this regard that the resolution is so much higher it also has more color information. You can pretty much make a 4:4:4 DV image out of HDV.

[Glenn Chan]

All DVDs are 4:2:0, so this 'completely flawed' system is standard world-wide.

There are actually different 4:2:0 schemes, and they make a difference.

The first difference is between co-sited and interstitially-sited chroma.
--For co-sited, linear/triangle resampling is usually used. It suffers from generation loss, but has reasonably good quality.
--For interstitially-sited chroma, box resampling is usually used. It doesn't suffer from generation loss. In NLEs, it will recompress 1st generation material without generation loss.
So if you have a cross dissolve, most NLEs will only render the dissolve and not both sides of it. So you see 1st generation --> 2nd --> 1st. If there were noticeable generation loss, that would be a bad thing since you see a jarring jump from 1st to 2nd to 1st generation.
Box resampling doesn't suffer from that. But it has very poor 1st generation quality. The chroma looks boxy, whereas the linear approach smooths the chroma out (kind of like blur).

In practice, things get screwed up.
The standards for 4:2:2 formats call for co-sited chroma horizontally. For MPEG-2, it's co-sited horizontally and interstitially-sited vertically. Most cameras will follow the standards.
Most NLEs and compression codecs do not follow the standards- they tend to use something like box resampling (or nearest neighbour... which is worse) for both dimensions.
Most decoders (on the viewer's end) will assume co-sited chroma and do linear resampling.
So you can end up with a mismash of chroma resampling schemes, effectively causing the chroma to get shifted 0.5 (or ~1.0) pixels in some cases. And you can get the poor quality of box resampling (on either its encoding or decoding side).

Though these artifacts are not too bad.

2- As Thomas touches upon, interlaced images screw things up. The long explanation is in here:
http://www.hometheaterhifi.com/volum...ug-4-2001.html
(That article took me some time to figure out / understand.)

The oversimplified answer is that interlaced 4:2:0 effectively performs similarly to 4:1:1.

3- What you see on your NLE's preview screen is not what you get!!

NLE codecs tend to use box resampling, which looks terrible and passes the chroma through (for that particular chroma scheme format). The viewer's decoder will likely use linear resampling, where the chroma look better.

So encode to MPEG2 from your NLE, and use a DVD player like powerDVD (software) or a hardware DVD player (ones where you stick a physical DVD into them). This is a somewhat better representation of what the viewer will see (*though you can run into scaling artifacts if the display doesn't do 1:1 pixels).

4- Chroma subsampling artifacts get worse the more saturated your colors are (saturated as in more pure). Most real world scenes don't contain extremely saturated colors and you don't really have a problem. Titles can be problematic.

4b- The 4:1:1 and interlaced 4:2:0 schemes are too much chroma subsampling IMO... and it's also inefficient compression. Preferably you would avoid those schemes. Though you can get pretty good results from those formats, given enough time in post (e.g. the "carnival of rust" music video looks amazing and was shot on HDV).

[Peter Moretti]

So does this mean that 24P works better with 4:2:0 than 60i does?

[Glenn Chan]

So does this mean that 24P works better with 4:2:0 than 60i does?

Yes, as long as you're using a progressive chroma scheme.

In the HV20, this may or may not be happening. AFAIK the HV20 doesn't record 24p material without adding pulldown to it. The result is a 60i stream... where you should be using an interlaced chroma scheme.

Not all Hollywood releases are actually made that way. They might be put onto a 60i format and they may not bother removing the pulldown.

2- Granted, it's hard to tell what's going on since this JPEG has its own 4:2:0 chroma subsampling. (Decoding of that chroma differs between codecs... e.g. photoshop and firefox give different results.)