BBC HD

I think both types of comparison are valid. If you said "HD looks X times better than SD" but 90% of HD content was 25Hz and 90% of SD content was 50Hz and would give twice as smooth motion (or even seemingly better motion because of screen size/viewing distance), I think it's valid, since in that case, most broadcast SD content would give a lot more accurate motion than most broadcast HD content.

I think it's sort of - what quality on average can you expect from HD and what quality on average can you expect on from SD. You could also do figures for the best of each.

Comparing 50i (50Hz) SD with 25p HD?
Is that how all the best de-interlacers work in all the best current HDTVs? Basically change the part in motion to line doubling? Don't any do motion compensation on the parts in motion (I'm not talking about if the interpolation feature is on) or anything better than switching the whole frame to line doubling when everything is in motion?

But even if the resolution reduced by half in the 50Hz version, it would still be easier for the eyes to track - the 25Hz one would have 'judder' and not be as easy for the eyes to track (eg. would sort of blur - and not just motion blur).

From Digitalspy,

Have a look at this courtesy of mwardy, one cap is takin off BBC One HD boadcasting at 1440x1080i the other the same moment on BBC HD at 1920x1080i

http://dl.dropbox.com/u/5159779/Wimbledon.htm

Look at the net in particular.

That could easily have far more to do with a higher bitrate, or even whether it was an I/P/B frame of either broadcast. Gasquet is blockier in the worse image.

If you resize the 1920 image to 1440 and then back up again, the difference between it and the original 1920 picture is far less than the difference between the 1440 broadcast and the 1920 broadcast.

Minor point, but those also aren't exactly the original broadcast images - they've been deinterlaced.

David

Why has the 1920x1080 one chopped off a few lines at the bottom, added a few black lines at the top, and moved everything down?

At a guess, it's just one of those things in the chain - it's two pixels because it's a single pixel per field.

Here's a less wobbly version, but it doesn't work in IE (cos it's rubbish) - it works in Firefox though, and probably Chrome, etc.

linky

David

It looks even more obvious in your version in your link, I wonder how those screenshots have been achieved

Actually, it doesn't work because your html isn't valid, and it doesn't know for sure how to render it, so it goes into "quirks" (aka pretend I'm IE6) mode. You can override that by providing a valid structure and doctype, or by (quicker/easier) adding this at the beginning of your file:

<meta http-equiv="x-ua-compatible" content="IE=8">


</offtopic>

Strictly speaking yes, but when I did give it a valid doctype it kept flashing to white when I hovered or unhovered, which made it useless for the desired purpose, so I left it at the simplified version

David

Works fine here...

But increasingly SD broadcasts are SD downconversion of HD originations. If it's shot 25p HD it's 25p on HD and SD. If it's shot 50i HD then it's 50i on HD and SD.

If you want to compare 50i and 25p then that's fine. If you want to compare SD and HD that's fine. But comparing 50i SD with 25p HD is not a uniform comparison.

50i and 25p are aesthetic choices made by a producer or director. HD origination may have proportionately more 25p origination than SD these days because the high-end productions that transitioned from SD to HD first were often 25p SD productions (or 25p Super 16mm productions) - like drama and high-end documentary. Sport and entertainment have also transitioned though, and they are still 50i at HD. The remaining SD-only origination is more reality, news and daytime dominated, and they are bastions of 50i SD content.

So whilst I agree with you that an HD originated show may be, currently, more likely to be 25p than an SD originated show, I don't think you can say HD=25p and SD=50i when making an "SD vs HD" argument without qualification.


Hmm - still not sure I agree. Take NHK World for instance that's close to 100% 60i originated.

I think when comparing HD with SD you have to compare HD with the same content downconverted to SD - i.e. compare watching EastEnders in HD with watching it in SD, or compare watching Cranford in HD with watching Cranford in SD.


It's how the interlaced signal works.

Fast moving content is captured effectively at 720x288/50p - as there is significant motion between fields. That's what comes out of the camera. It's how interlaced systems work. On static content you get 720x576 detail (assuming no prefiltering vertically...) and on fast motion you get 720x288 detail. Between the two you get something in between due to spectrum folding.



Some de-interlacing algorithms will use phase correlation (Snell's Quasar) and motion compensation, and some will use vector adaptive techniques, but all of these are effectively guesstimating (they can't do anything else) - as there is no guaranteed information in the source to deliver this information. It's how interlacing delivers bandwith reduction. You can't have your cake and eat it! You have to decide how to interpret information that could generate an identical interlaced signal either from fine vertical detail or vertical motion. Looking across multiple fields can guide this process - but it can't make it guaranteed to be right. There are usually ways of generating content that will defeat de-interlacers.

You have spectrum folding of vertical motion and vertical detail in interlacing - this is unavoidable. To avoid interline twitter (where fine vertical detail gets folded into vertical temporal information and appears to flicker at FRAME rate) there is some pre-filtering (like frame-line-offset line-averaging in 1080/4320p based cameras) for instance.

If you are comparing 1080/50i (which degrades to 540p at high speed) with 576/50i (which degrades to 288p at high speed) you are again comparing like with like, whereas if you compare 1080/25p (which could deliver full 1080p resolution on fast motion) with 576/50i (which degrades to 288p at fast motion) there is a significant difference.

Of course a 25p capture with a 1/50th shutter would have significant motion blur, but you could chose to use a different shutter... Doing the same at 576/50i won't help you get around the degradation to 288p...

But choosing a different shutter for 25p that reduced the blur would only make it judder more - and 25p already judders a lot more than 50Hz content.

Does that mean "they blur it a little bit"?

David