AG-HVX200 vs GY-HD100U for Adventure Sports

[Uri Blumenthal]

Originally Posted by Barry Green

This isn't digital uprezzing we're talking about here, this is a valid resolution enhancement technique that's used by just about all the major manufacturers.

No question that it's being used by other vendors - though not to such a drastic extent, AFAIK.

It is a way of scanning the CCDs to enhance resolution. You have to keep in mind that the CCDs are analog devices, not digital, and they are scanned into YUV space, not RGB. It's not pixel for pixel. What happens on any particular CCD pixel bears no direct relationship to what happens in the scanned frame's pixel; each YUV pixel is made up of about 60% from the green CCD, 30% from the red, and about 10% from the blue. It is the way and pattern that these pixels are combined that delivers the final YUV image.

I sorta understand, and agree that there is no direct relationship between CCD number of pixels and the "resulting pixels". But surely there is a relationship, and all other factors being equal quality of result is in direct proportion to the number of CCD pixels. Regardless of how you mix the CCD pixels together, the less pixels you got in the beginning - the less "stuff" you have to play with.

Or are you saying that the number of pixels on CCD does not matter? And the up-scaling algorithms are good enough to make scanned frames practically indistinguishable between say "native" 1270x720p and "up-scaled" 960x540?

In many ways pixel shift is akin to bayer demosaic'ing................It has also been vetted by the BBC, who determined that it is a perfectly valid technique that (on the surface) claims as much as a 50% increase in resolution over the nominal chip count, but results in a more practical maximum of 1.414x as much res, and as far as the BBC is concerned they think a 33% boost is about the most that the technique can legitimately deliver.

33% boost would roughly be a step from 1440 to 1920, or from 960 to 1270. Does this BBC technique condone the simultaneous "boost" along both axes?

Boosting from 960x540 native chip count. If it's 33% of the total number of pixels (as I suspect it is), then you can boost 960x540 to only about 957x720. If we take more liberal 50% boost - we still end up with only 1080x720 of "legitimate" (according to BBC findings) pixels.

And for HVX200 to "always be 1920x1080 internally" you need doubling (i.e. 100%) on both axes.

As a side-remark: if pixel-shifting was that great - why would Canon spend extra and change from "pixel-shifted" XL-1S to "full-size" XL2?

[Chris Hurd]

Originally Posted by Uri Blumenthal

No question that it's being used by other vendors - though not to such a drastic extent, AFAIK.

The trouble with your statement is the "AFAIK." What you may not know is that Pixel Shift in both H and V axes is not at all uncommon in both standard and high definition video cameras..

Does this BBC technique condone the simultaneous "boost" along both axes?

It's never been a problem with them before. Pixel Shift in both axes has been around for quite awhile... Panasonic invented it more than a decade ago. It has *never* been an issue for *any* broadcaster.

As a side-remark: if pixel-shifting was that great - why would Canon spend extra and change from "pixel-shifted" XL-1S to "full-size" XL2?

Actually the XL2 uses Pixel Shift also, in the H-axis. Despite your error, I'm glad you've mentioned the Canon XL. It's one of my favorite counterpoints to the fallacy of believing that something must be wrong with the HVX because it uses double-axis Pixel Shift. The fact is that the Canon XL1 and XL1S both use double-axis Pixel Shift to achieve DV resolution from CCDs that were less than standard definition in pixel count. It was never a big deal then, so suddenly it's supposed to be a big deal now? Sorry, but no. The track record of double-axis Pixel Shift was proven with the original Canon XL1 (to name only one of a number of examples). Nobody complained then. I don't see how anyone can complain about it now. Sd, Hd, it doesn't matter... the process is the same.

I have no idea why this thread was ressurected when its most recent reply was almost a month ago... and before that, since June. Barry has summed it up quite well: the proof is in the image. I would like to encourage the measurebators and naysayers to please get your heads out of the numbers and look at the screen.

All that counts, is how good it looks. Thread closed.