1080i vs 720p

P

Phrank

Member
According to The HVX Book, "The choice between 720p and 1080i/p can hinge on a few factors,..." The book goes on to describe these factors as ultimate frame resolution, variable frame rate requirements, recording time limitations and destination network format requirements.

Since the books publishing, are there any newly discovered factors related to picture quality that one might consider before choosing between 720p and 1080i for D5 delivery of raw footage, all other factors aside?

If there are any other threads covering this topic, please direct me!
 
It's still under debate to some extent. Different people report different things as far as how much better 1080p is than 720p. Barry initially said he say an almost 30% improvment with the pre-release model though last time he mentioned it he said 20% (don't know if this number changed based on the release model, further testing, or what). It also depends on what you are shooting. When I shoot something REALLY designed to push the resolution (like the surface of an old brick) then I've noticed a good 25% increase, maybe more. On the other hand, for lots of other shots I don't see much of a difference at all. It really just depends.
 
If I'm not planning any variable frame rate shots I shoot 1080. Otherwise it's 720P all the way with some FX shots in 1080 for better composites.
 
I really think it should also be noted what you are planning to finish as. Most of my stuff is going to an SD DVD and, in that case, 720p is just as good as 1080p. Even if you're working for an HD broadcast you should really consider what people will be watching this on. Most people with HD TVs have plasmas that don't even support full 720p resolution. Some networks also choise 720p (which others go 1080i). Personally, with the benafits in card space, I really wouldn't bother with 1080p unless:

a. I was planning either a film out or some other form or projection.
b. I was making a 1080p Blue Ray or HD DVD disc.
c. I was shooting for a network that either broadcast in 1080i or required it.
 
Could a credible argument be made that the HVX is a 960x540 CCD sensor, using pixel shift to get 1080. Thus the resolution of 720P is better than 1080i from this camera?
 
No, not at all -- just look at the footage, there's more resolution in 1080 mode than there is in 720.

The argument is that the HVX scans a 1920x1080 image off its sensor, which results in about 1400x810 of resolved detail. That gets processed internally at 1920x1080, and then when it gets recorded it gets scaled to fit the recording format.

For 1080, that means 1280x1080 (which is almost enough room to retain all the resolution the imaging head is capable of).

For 720, that means 960x720. Which is a lot less than the original 1400x810.

So no, I wouldn't say that it's better 720 camera. The appeal of 720 is the "native" frame rates, and the footage looks awfully good anyway, so a lot of people consider that the "sweet spot" of the HVX -- a little less res in exchange for 2.5 times as much recording space. But if you want the most res you choose 1080p mode.
 
^
DVCProHD scans 720P at 960x720 and 1080P at 1280x1080. The argument is that the eye is less sensative to horizontal resolution so Pansonic is willing to drop more of it (as opposed to the 1440x1080 for 1080p for most recording methods) in exchange for no motion compression and more color space (compared to the 4:2:0 or HDV or 3:1:1 of HD Cam I believe).

Our curiosity Barry, you quote the camera as having resolved detail of 1400x810 which is theoretically what the pixel shifted 960x540 CCDs should yield.

At the same time we see reports from the intial shot out with Adam Wilt claiming the HVX200 only produced 550 lines horizontally and 540 vertically. The vertical number would indicate that the pixel shift didn't add any rez vertically and the 550 horizontal claim...is just a bit confusing since the CCD is 960 and capture 960 in 720p without any pixal shifting. He also claims the JVC at 700 lines in both directions. At the same time, I've seen some of your side by side comparisons of the HD100 and HVX200 showing clearly to my eye that the HVX has more vertical resolution than the HD100 and the HD100 has a bit (but not much more) horizontal resolution. It certainly didn't gel with the 550x540 vs 700x700 numbers we got from the shoot out. Further confusion is caused from Shannon Rawls saying he "didn't see 700 lines from the HD100."

I realize how silly all this pixel counting can get but I am a bit curious why there is such a discrepency between different assesments, especially by those who were doing the same tests. Is it just a matter of perception when judging these things and people disagree with what a line or resolution is (so you give the nod to the HVX slightly, Adam to the HD100 by a lot, and Shanon to the HD100 by a tiny bit though he says the 700 number is off) or is there something else I'm missing?

Also, why exactly do the cameras have trouble resolving anywhere close to their horizontal resolution? (HD100 doing 700 at best with a 1280 chip and 1280 sampling, HVX200 doing 550 at worst with a 960-1400 chip and 1280 sampling)
 
DVCPro HD records 720p in 960x720 with non square pixels. The 960x720 frame is stretched on playback to fit 1280x720.

It's basically an anamorphic process, like many other things in our field.

EDIT: David beat me to it with a lot more info. :)
 
DavidBeier said:
^
DVCProHD scans 720P at 960x720 and 1080P at 1280x1080. The argument is that the eye is less sensative to horizontal resolution so Pansonic is willing to drop more of it (as opposed to the 1440x1080 for 1080p for most recording methods) in exchange for no motion compression and more color space (compared to the 4:2:0 or HDV or 3:1:1 of HD Cam I believe).
Click to expand...

well, it IS true that more color space and less compression artifacts are more important to me that straight res. That's why I had to rule out the HD100.

Thanks
 
You know, since I've had the HVX I've been meaning to do a good test comparing the actual differences between recorded 1080 and 720, but had never got around to it until today. It's actually very easy to do, but other than one early attempt that was marred by different framing and exposure between the two sides, I've not seen it done before...

The test... locked off camera, recording a short clip in 1080 24PA and the same at 720 24PN. Same exposure between them, exactly 5.6 on the iris, pretty much the sweet spot. Focus is set to the text near bottom of screen, right at the "split" between sides.

The difficulty in doing a test like this is that you are FORCED to scale one or the other or both images before examination. I didn't trust letting either Raylight or CineForm do that for me, so I just used the raw .MXF's which will load in native size and aspect when the P2 Viewer software is installed.

I loaded both clips into Digital Fusion and used Catmull-Rom scaling on both, taking them either to full raster 1080 or full raster 720. Finally, I did an extreme levels test to bring out the compression artifacts in both of them.

The results? No question that the HVX recorded 1080 has a slight edge on resolution. How much sharper? This test can't address this, but I'd say 20% might be a safe estimate. The glass half empty view is that the HVX is an extremely tepid 1080 camera, the glass half full view is that you can actually exceed the recorded resolution of the Varicam! Is this bit of extra resolution worth the 2.5X increase in P2 bandwidth when recording? Your decision on that one. I'd say for wide-shots or VFX then absolutely. The story isn't over...

The real shocker is how MUCH more compression is visible in the 720 original recording... much, much worse than the 1080 stuff. This "proves" the conventional theory that blue screen and green screen will be easier and better with 1080 original, and actually is the stronger case for shooting 1080 over the resolution issue, IMO...

Also, these tests help validate the theory that shooting 1080, then scaling to 720 full raster for editing with CineForm is a very good workflow, as you keep bandwidth in line for 720 material, get the resolution "boost" of the 1080, while still keeping the compression artifacts low...

Here are PNG's of the three tests in this directory...

http://ftp.datausa.com/imageshoppe/outgoing/HVX_TESTS/STILLS/

Regards
 
Great stuff Jim. In my own tests, while I noticed a solid difference in 1080p, I didn't see any advantages to 1080p downrezed compared to native 720p. Clearly I need better downrezing software.
 
DavidBeier said:
...in exchange for no motion compression and more color space (compared to the 4:2:0 or HDV or 3:1:1 of HD Cam I believe).
Click to expand...
Not to hijack the thread, but I always thought that color space was expressed as a ratio so wouldn't 4:1:1 be less than 3:1:1?

-A
 
DavidBeier said:
Great stuff Jim. In my own tests, while I noticed a solid difference in 1080p, I didn't see any advantages to 1080p downrezed compared to native 720p. Clearly I need better downrezing software.
Click to expand...

Thanks David. I learned today that the native differences between the two resolutions are actually so slight that a bad resizing method can completely change the outcome of the test... I kept apples to apples, but if I was to use "sinc" resizing on the 720, it will look every bit as sharp to my eyes as the 1080 using Catmull-Rom, and using bi-linear on the 1080 will make it look actually worse than the 720.

The difference in compression is a whole 'nuther thing, and the true benefit of shooting 1080. My new theory is that this would even be apparent on a panning shot, where sharp lines and details moving between macroblocks should be better at 1080 than at 720...

Bottom line, unless you have to have slow-motion, DON'T SHOOT GREENSCREEN AT 720P on the HVX. Maybe I should put that in caps... :)

Regards,
 
Thanks Patryk... I've got a few more ideas I hope to test out soon...

Right now here's something very revealing about the difference in compression artifacts between 1080 and 720p, an animation of my above example still... This QT is a cropped extraction from a 1080 comp, with 1080 source on left and 720 source on right with the split screen going between them, heavy levels, and the footage is running... notice the amazing difference between them, and how the boarders between colors are so much sharper in 1080...

http://ftp.datausa.com/imageshoppe/outgoing/HVX_TESTS/720_wipeTo_1080.mov

15megs, please don't stream, but "save as"...
 
Jim...

Thanks for the tests and for posting samples. Especially telling was the .mov file with the wipe.

I recently shot a commercial which required extensive green scree work and used 1080 p30. Was easy to key, and even flying hair held up in the matte.
 
You must log in or register to reply here.
Back
Top