HMC150 720p vs. 1080 resized to 720

B

Barry_Green

Moderator
A question I've been thinking about is -- we know that the HMC150 can deliver about 1440 x 810 worth of resolution, which means that 720p should be close to capable of storing all the detail the camera can generate, right?

So why not just shoot in 720p mode?

And the thing that's nagging me is the color sampling. With the HMC150's AVCHD format, the frame gets resized and then the color gets cut to 1/4 the resolution (using 4:2:0 color sampling, like HDV or XDCAM-EX or XDCAM-HD). Which means that on a 720p frame, you get 1280x720 of brightness, and 640x360 of color. Whereas with 1080p mode, youll have a 1920x1080 frame, which may only have 1440x810 worth of detail, sure, but -- the chroma sampling will be at 1/4 the res of the full frame, meaning 960x540 of chroma. Well, that's a big difference, right? 640x360 = 230,400 chroma samples, vs. 960x540 = 518,400 chroma samples, or about 2.25x as much chroma...

The question then becomes: would you gain much by shooting in 1080 mode and downconverting to 720? You wouldn't gain much of anything in terms of overall image resolution, but you'd more than double your chroma res... So, does it make a difference? Um, well, yeah.

So I shot the exact same image in 1080p and 720p mode, and put the files side by side on a 1280x720 Vegas timeline, and copied the frame and pasted to PhotoShop to compare. This image is magnified to 400% to show the difference, but obviously, yes, downconverting in Vegas from a 1080p source is going to result in much higher chroma resolution.

Look at the black & white stuff, it's almost identical, but in the red and in the red/green transition, the downconverted side is just way more detailed.

AVCHD-1080-resized-vs-720-native.jpg


So, when can you use this? Should you *always* shoot in 1080 mode? Not necessarily; remember that if you're outputting to 720p Blu Ray or something like that, it's going to get chopped back down to 4:2:0 anyway, giving back the advantage you tried to gain. But if you're outputting to uncompressed or to AVC-Intra or MPEG-422 or some other 4:2:2 codec, yes there's certainly quite a bit of color resolution that can be gained back. If you're outputting to 1080, there's nothing to be gained here vs. just shooting 1080 native and keeping it that way, of course.
 
Hmmm... so there is something to be gained by shooting 1080 on this camera. Barry, do you know not consider the 720 the sweet spot anymore?
 
I'm trying to get to the bottom of that... I'd really have to try some complex shots to see if 1080 pushes the compression harder than 720. I mean, obviously, on the surface it seems like it should -- twice the pixels, same bandwidth, = more compression. BUT -- if the overall amount of detail is about the same, even though it's on twice as big a frame, then AVC is such an intelligent codec that I think it might perform better than we'd expect.

I sure wish I had two of these cameras to do some side-by-side, simultaneous shot comparisons, one in 1080/24, the other in 720/24...

I like 720 on the HMC150 because it's a lot faster to edit, and it retains almost all the (luma) detail the camera can generate, and (at least theoretically) it should be the least compressed. But, nothing beats actually just opening it up and testing it, and now I'm starting to wonder... there's a pretty big chroma resolution advantage to 1080 mode, no question about it. If the overall image holds up equally well, then... yeah, I might switch over to saying "shoot 1080 all the time" unless (and this is a big "unless") you need the "reality" look; I'd definitely choose 720/60p over 1080/60i any day.
 
great. thanks as always for your knowledgable info Barry! as per my other post, how well does the hmc150 image hold up on a 70" LCD/plasma(especially comparing 720 to 1080p? i will be eyeing this thread closely.
 
Last edited:
Barry this is crazy! I thought about doing this same test last night after reading a couple of post about the difference between 720 versus 1080 with the 150. I did not think so much about how much chroma resolution you could gain by shooting 1080 over 720.

What I really would like to see is if there is any noticeable compression difference when shooting motion in 1080 and then downconverting to 720, versus shooting motion in 720.

We all know that the compression is the same between the two resolutions, but what we don't know is how the final 1080 image will hold up compared to 720 if it is downconverted afterward. Any ideas??
 
Barry_Green said:
A question I've been thinking about is -- we know that the HMC150 can deliver about 1440 x 810 worth of resolution, which means that 720p should be close to capable of storing all the detail the camera can generate, right?
Click to expand...

Sorry, I haven't been following the hmc150 closely enough, why do we know that it can deliver only 1440x810 worth of resolution?
 
Al MacLeod said:
I imagine you did this type of comparison with the hvx200...do you have a link to it?
Thanks...
Click to expand...
No, with the HVX it's 4:2:2 vs. 4:2:2, so there's already so much more chroma info to begin with that the difference isn't that much. There's already chroma on every line, so you wouldn't gain that much on the HVX. Plus, with the HVX 720p starts at 960 anyway, so I've always recommended (and continue to recommend) that anyone wanting the best out of the HVX should be shooting in 1080p mode, no questions asked.

I thought the situation might be reversed with the HMC150, but now I'm not so sure. More testing about codec resiliency is warranted. If the 1080 mode is equally resilient to compression artifacts vs. the 720 mode, then yeah, 1080 mode would be what I would always want to use.
 
ecking said:
Sorry, I haven't been following the hmc150 closely enough, why do we know that it can deliver only 1440x810 worth of resolution?
Click to expand...
That's Panasonic's numbers, based on their white paper for the HVX200 about Spatial Offset ("pixel shift"). They claim that an HVX is capable of maximum observable res of 1440x810, and the HMC150 uses the same basic technology so I'm extrapolating those numbers forward to the HMC150.
 
From my non side by side tests,
I concluded that you will retain more color if shot in 1080 if:

You have full light (shooting with more than wide open lighting).
You are shooting progressive.
Your final output will be 1080.- EDIT, actually downconverted will also look better, same as doing stills in photoshop
You edit in a different codec than AVCHD.

Otherwise shoot 720/24p if:
You need to edit quickly.
You want to stay in AVCHD
You are in low light, as there will not be much color difference anyway.
It's not a once in a lifetime type of footage, or my own stuff.

For me it is more about what I am doing with the footage than anything else.

720p is so much easier to edit, 720p is so much easier to edit, 720p is so much easier to edit.

Did I say how easy to edit 720p is?

...Anyway, I have also been shooting with another brand camera, that does not do 720p, so I may also shoot 1080 due to my workflow in converting footage to 1280x720 on the other camera, at the same time I convert to cineform both from 1080 to 720/24p

Or edit raw 1080 cineform when I want better color, or do any filters/ grading.
 
Last edited:
What about 1080 50i vs 720p 50?

I shoot documentaries with some motion and some pans, so 25P stutters far to much, thats why i shoot in 50i or 720p 50. I know that many people avoid interlaced footage like a plague, but why? Please, enlighten me! :)

/Roger
 
My main reason was interlaced artifacts or that combed look around the edges.
I hated that because it was a constant distraction to me even when no one else complained about it.
Then I would spend time de-interlacing which always made the images a little softer.


I've seen interlaced footage that looked out of this world but I could never get it myself.


If works, it works - and I think interlaced versus progressive in a non-broadcast workflow is more of a personal choice but when shooting for some stations you will have to shoot interlaced.

Me. I just think progressive is the way it should be because the image is not so fractured.
 
Here's a question: if your final output is 1080P (say, Bluray), but you want to use 60P for frame accurate slowmotion, how well would 720 native footage upconverted to 1080 blend with the native 1080?

Something I think I am going to try experimenting with before I begin my upcoming project, as I definitely wanted to use some 60P for slow motion...
 
Cool test Barry. 1080 is getting a lot of push here recently. The chroma argument always made sense, although someone did make a good point about pixel shifting not giving you the same boost in chroma as in luma. The concept was that to get the higher luma rez, the pixels are offset from the R/G/B chips... however, you don't get the same gain in chroma since you can't gain red detail from the green or blue chip, etc. Regardless, if the chips can pull 960x540 of chroma, there is quite a bit being missed from 720p only storing 640x360. Even if you are outputting to 720p 4:2:0, that would be enough reason to shoot 1080p if you plan to do CC work.

One factor that flaws any tests between the two is scaling. The real-time down conversion from 1080 to 720p in-camera is probably not the same algorithm used by your software (and software gives you several options, some which work better for certain material). Oddly enough, I noticed some circumstances where 720 unscaled to 1080 looked slightly better than the same shot at 1080 (notably hard edges and flat colors). However, I don't think that is the norm, and you should be able to apply the same processing on the 1080 pic to get the same or better results.

Recently I was testing 1080 with highest shutter speed with full zoom and quickly panning around the neighborhood, shaking the camera and doing fast full zooms. I am amazed at how sharp the individual frames hold out!!! In the worst possible conditions, edges tend to soften rather than block up which is what your eye sees at normal speeds anyway. I believe that in most cases, the codec can deliver far more than this camera's chips are capable of. I am wondering if we will soon see a "3MOS" camera released at the same level that will really test PH mode (just don't do any of those fast zoomed pans).

The original reason to stay in 720 was for the least compression artifacts. Except in the worst conditions, it doesn't seem to be as much of a valid reason NOT to shoot 1080. It makes it even more unfortunate the lower bitrate modes don't shoot progressive! I still think 720p60 is one of the best features, but have been sticking with 1080p for other framerates.
 
Jim H said:
Here's a question: if your final output is 1080P (say, Bluray), but you want to use 60P for frame accurate slowmotion, how well would 720 native footage upconverted to 1080 blend with the native 1080?
Click to expand...
Since the difference between the two is not very drastic, it will mix well. The awesomeness of the smooth motion will take any attention away from the small difference in quality... seriously! Just don't use the DVCProHD converter, it cuts a lot of the resolution out of 720p.
 
mcsmooth said:
1080 is getting a lot of push here recently. The chroma argument always made sense, although someone did make a good point about pixel shifting not giving you the same boost in chroma as in luma. The concept was that to get the higher luma rez, the pixels are offset from the R/G/B chips... however, you don't get the same gain in chroma since you can't gain red detail from the green or blue chip, etc.
Click to expand...
Whether that argument holds in the general spatial-offset discussion or not, recognize that it's pretty irrelevant in the discussion here because the 1080 mode of AVCHD stores 960x540 chroma anyway, so you should certainly be getting maximum advantage of the chips' resolution by recording in 1080 mode.

In the worst possible conditions, edges tend to soften rather than block up which is what your eye sees at normal speeds anyway.
Click to expand...
AVC is designed specifically to do that; instead of using big macroblocks like MPEG-2, it uses scalable blocks that range from, I believe, 16x16 on down to 4x4, so it can have much smoother transitions without the big blockiness.

I believe that in most cases, the codec can deliver far more than this camera's chips are capable of.
Click to expand...
I'm sure that's the case. I'm actually interested in trying out the new HMR10 AVCCAM HD-SDI recorder; I would love to put PH mode up against XDCAM-EX from the same source to see if my prediction (21mbps PH ~= 35mbps MPEG-2) is accurate.

I am wondering if we will soon see a "3MOS" camera released at the same level that will really test PH mode (just don't do any of those fast zoomed pans).
Click to expand...
No idea, but the pieces are there; an HMC200 or 250 or whatever, with the HPX300's chipset, would give you the sensor resolution to fully tax the codec. No idea whether something like that is forthcoming, but if it was, it'd push AVCHD harder. Of course, then you'd have cmos issues to deal with, but there's no free lunch.

I still think 720p60 is one of the best features, but have been sticking with 1080p for other framerates.
Click to expand...
Agreed on 720/60, still evaluating on the 1080/24p vs. 720/24p, but if it holds up in torture testing then yeah, I'd agree that 1080 would be the only way to go.
 
so for greenscreen work, 1080 would be a good idea?

while for regular with no concern for colors, 720p would be fine?
 
You must log in or register to reply here.
Back
Top