1080p compared to 720p

[Anders Holck]

I have made a small theoretical test, using a shot from Kaku's server.

Basically I wanted to get an indication if there is enough detail in the 1080 mode, to make it justified over the 720p mode. And would it be possible to inter-cut both formats

My procedure was totake a frame from the native DVCPRO HD clip both upress' one of the original 1080p frames to full 1920 x 1080 raster and also also convert the same frame to 720p DVCPRO HD in FCP 5 and up-convert that to 1920x1080. Then take both into Photoshop and compare them.

This will give a hint of what advantage there will be to 1080p, resolution wise.

You can download a layered .PSD file here:
www.andersholck.com/1080_compare.psd
And do the comparison yourself

My observations:
It's obvious that the resolution is pretty close, but you can see a definite improvement of the 1080p in the fine metal structure in the Iron tower in the upper middle.
Also a little better, are the numbers on the numberplate. But I think it's very close.

What is missing in this image is clearly some strong color details that would illustrate the better chroma resolution of 1080p.

Edit: As some have commented, this is of cause a theoretical test. If done correctly you should actually shoot the same scene in both 1080p and 720p, as differences in CCD readout, dsp algorithms and compression could skew the result. I did this test to get a hint where the 1080p mode is resolution wise and see if it would boast a noticeable improvement over DVCPRO 720p. And as I don't have the camera this was the only way I could possibly do the test. Still I think it's an interesting test especially as we often discuss luma and chroma sampling on these boards, and it's interesting to see how it affects real world imagery captured by the real CCD.

[Darkline]

wow, thanks for that Anders.

For me, this shows there is no advantage to shooting 1080p. It really is so little (at least on this test)

I guess because the capture is 1080, then downsized to 720p, this would make the difference even smaller. However it's beginning to become clear that 1080p mode is not higher than 720p. I wouldn't even say there was a 5% increase in resolution here.

I'm staying open minded though; I look forward to more tests.

[mmm]

In some ways I am pleased if 720p isn't much better as it makes P2 cards a more realistic proposition. If I can fit 16 mins or so with 2 4GB cards, that isn't too bad (shooting at 720p25)

[Anders Holck]

Of cause it also depends on the resize algorithm of the HVX-200 DSP. In my test I have used "normal" in FCP 5 which should be equal to a Bicubic resize.

[smelni]

I am a little lost as to what the steps of the test were.

The two images are similar but there is striking differences in small details (license plate for example) - if this image were projected on a very large screen then it might make a difference.

But again - what is being tested here - A resize or the image acquisition - if its not the image acq then is it a valid test?

[Rodrigo]

"5% increase"

HAHAHAHAHA... what does that mean? you mean 5% more pixel count? 5% more vertical resolution? 5% more "perceived" resolution? you did it by eye?

Anyway, the test is flawed from the beginning: for that u should compare two footages from the same camera, same settings, same position: one in 720p amd the other one in 1080p. the camera will not be real rez in 720p either so I bet the differences will be noticeably bigger. Anyway I perceive a real increase in resolution even in this test, a resolution that can be very useful for filmout

[Anders Holck]

What I wanted to test here was if there really is more detail in a HVX-200 1080p image than could be contained by the DVCPRO 720p codec.

I took a file which was originated in DVCPRO 1080p and recompressed it into 720p, then upressed both files into full 1920x1080 to compare the subtle differences.

You could do the same test by shooting the same subject twice, one in 720p and one in 1080p, then do the same comparison. But unfortunately I don't have the camera.

What could cause errors in this test is that the clip is recompressed again in DVCPRO codec degrading the 720p image a bit. Also the quality of the downconversion could be better in FCP than it is in the camera.

In the HVX, both 1080p and 720p originates from the same resolution CCD image, so this test, while not being 100% correct, should give a pretty good hint about the difference. And if there's an error, it's bound to make the 720p footage look worse, because of the second compression pass, especially as I think Panassonic's hardware encoder is better than Apples encoder.

[Anders Holck]

Here is another test. Another clip, same procedure.

www.andersholck.com/1080_Compare2.psd.zip

This time there is four layers, two of them with interpolated chroma channels to make up for apple's non smoothing codec.

Still the resolution is a little better in 1080p, and this time the more saturated colors in the red lights show a lot more chroma resolution.
Also on the back of the silver truck in the left side there is a sign that reads "Transport Communication Sagawa". Here "Communication" is much sharper in 1080p.

[smelni]

again - i am not sure if this is a valid test - although both images originate from the same CCD in camera there is a whole lot of other processing that can occur by the time the codec compresses the image and that processing can be quite different for 1080 and 720.

Maybe pixel shifting is used differently. Maybe different thresholds are used for individual pixels.

The only thing this test is testing is the abilities of the codec. Not that that isnt useful so its a fun test to look at.

thanks for doing it.

[Anders Holck]

Yes, there could definitely be a lot of difference in the process path, I agree.

But I still think the test is valid as to where the 1080p mode is resolution wise.
It is hard to see it in normal footage, but its obviously below the 1280x1080, but also delivers some detail above the theoretical limit of 720p DVCPRO.

[Stevet]

Thanks Holck !

[mmm]

The test is far from perfect, but Anders isn't claiming that it is. However it does suggest that the HVX captures more detail in 1080p than in 720p. How much more can't really be judged because of the other factors involved, but it looks like it may not be a great deal.

[Darkline]

lol... yeh yeh..

5% of me being extremely subjective with no quantifiable mesurement to gague except my eye. I guess what Im trying to say is that it 'looks' almost unoticable, certinaly you have to look very closely to see any difference; which is a bit odd if it's meant to be twice the resolution.

But you illustrate the point perfectly. We need to see res charts being shot in both modes to get accurate comparisons. It's worth doing as we'll all save huge amounts of storage space if 1080p turns out to be a bid of a dud. I think its a mode still worth having as to my eye luma res doesnt look much higher, but chroma res could well be which would give much better footage for keying/comping.

[Shaw]

Yeah the chroma res does look good. I'd definitely shoot 1080p for compositing.

[Emanuel]

Thank you Anders for your contribution!

[Barry_Green]

However it's beginning to become clear that 1080p mode is not higher than 720p. I wouldn't even say there was a 5% increase in resolution here.
On a properly-shot test, there is a substantial difference. I'd say more on the order of 25-30% higher resolution for 1080p. I've shot resolution charts in both modes on the preproduction model and seen the difference.

Once I get a final released model I will conduct a similar test and post the results. But from my preliminary test I'd say 1080p is at least 25% higher res than 720p in the recorded footage.

[Antoine_Fabi]

Ah !!!

It's hard to keep editing and read good news like these every few minutes...

[PappasArts]

There should be 50% more res in the 1080 over 720 if Panasonic did it right.

[Stevet]

There should be 50% more res in the 1080 over 720 if Panasonic did it right.

Yes, technically, but I imagine Barry was adding in all the other elements that
make up perceived resolution.

Steve

[Barry_Green]

There should be 50% more res in the 1080 over 720 if Panasonic did it right.
That's not possible, Michael. Panasonic did "do it right", so let's not get involved in any brand issues.

The issues have been well documented for a long, long time -- the optics and the chips are combating the actual wavelength of light at this tiny 1/3" chip size.

There's only so much detail, only so much resolution, that can be resolved in a 1/3" chipset. So no, it's not possible to get double the resolution (a DVCPRO-HD 1080p frame has 2x as many pixels as a 720p frame). It is optically impossible to do so.

So they've designed the lens and the chips to work together to get the highest results they can.

[thisiswells]

"Pappas got run over by a reindeer..."

LOL.

[Mr. Blonde]

Isn't Mr. Papas the guy who got banned a while back for posting an article he wrote out to panasonic insisting on the hvx having an interchangable lens system?

[Darkline]

On a properly-shot test, there is a substantial difference. I'd say more on the order of 25-30% higher resolution for 1080p. I've shot resolution charts in both modes on the preproduction model and seen the difference.


this is really good news. When you're ready I'd love to see some examples/stills of the difference. The footage posted so far certainly does not seem to indicate it. But I trust your judgement so it's all good.

I've got 3 months to finish a worthy script before I even get my hands on one of these, so no rush here :-)

[Barry_Green]

I could show you grabs from the pre-pro model but that wouldn't be properly scientific. It should wait until I have a serial-numbered production model, otherwise I can't say that something may not have changed between when this one was made (probably five-six weeks ago) and the current in-the-customer's-hands model.

[Danilo Del Tufo]

Hey Barry, Panasonic still has not give to you a final model? It's strange!

p.s.: I've clearly understand that Chris Hurd email doesn't work, it's clearly an error of forum. This bad situation will go on and on till a day someone communicates this error in a clear form, writing it in a *hot thread*, so Chris could read, till that day I'm pretty sure that I cannot post there.
Did you have received my drawing? Did you like it, Barry? :)
If someone with *pure heart* can do this thing, I will be grateful to him.

Cheers :)

[Barry_Green]

Danilo can draw some Anime, that's for sure. Very nice drawing.

I forwarded your message to Chris. He's travelling right now and isn't really responding much at all. I think he's stuck on a dial-up connection and cannot do a lot of administrative things at DVInfo.

But he's been notified, so it's up to him to decide how to proceed.

[Emanuel]

I'm not a Danilo rep :grin: but thanks anyway for your concern. He deserves our care.

[Danilo Del Tufo]

Danilo can draw some Anime, that's for sure. Very nice drawing.



Thanks for the compliments! :embarasse :embarasse :embarasse
I'm not pointing to be a comic artist as profession, it's something that could be help for storyboarding, that's my idea!
As playin' classical baroque music with classical guitar to make my soundtrack! :) Or compose :)

I've already told in an old message that my teacher is a comic artist, he's very skilled, he is very talented artist Barry, you can find his arts here :

http://www.enzotroianocomics.com/

or simple typing "Enzo Troiano" in Google-> Images.

He has published on "HEAVY METAL" magazine in U.S.A years ago;) 1998.

http://www.heavymetalmagazinefanpage.com/0198.jpg

p.51-58 - "The Land Of Azor" - Troiano.

So he's famous :)

Last published work:

Korea 2145

Here You can find him at a convention : http://www.reggiocomics.it/protagonisti/troiano.htm


Cheers :)

p.s: my teacher will be happy that I speak of himself here, he doesn't know this community! :)

[Danilo Del Tufo]

I'm not a Danilo rep :grin: but thanks anyway for your concern. He deserves our care.

:embarasse :embarasse :embarasse :embarasse :embarasse :embarasse :embarasse :embarasse :embarasse :embarasse :embarasse :embarasse

thanks Emanuel...quando mi chiami? ti ho dato il numero di casa nel messaggio privato! /when do u call me? I gave u my home phone's number by private message:)