GH2 resolution in upsizing

Neverprod

Hi,

I did a small test to record in RAW 2.5K with the GH2 as the BMCC and i was surprised that it work perfectly, not because of the sensor but because the GH2 can offer a decent resolution with a awesome stability.

I was wondering what is the real limit of the GH2 in term of resolution ? (i'm talking about record video) ;)

I was thinking to make some footage with the GH2 in 2.5K Resolution, any risk ?

Can we leave this topic to discut about the resolution possibilities on the GH2 with hack ?

Thanks

Neverprod

Here a sample footage (i add a so small color grading)

Don't forget that this is a heavy compression MP4 with a bitrate of 20mb (not even)

mrbill

How did you set your gh2 up to shoot 2.5k? Thanks.

hedrox

you must have some kind of special gh2 when you can shoot 2.5k raw video with it..

thepalalias

@Neverprod It is really difficult to understand what you are saying since you do not talk about how you are setting things up or approaching this.

What do you mean, exactly?

Neverprod

The media info :) I set the GH2 in HD/Image Animation Change the 720p30p resolution to 2,5K 30p resolution

fatpig

it is upscaled. look for real detail. ;)

Neverprod

Sorry i deleted the video, i made a mistake (keep the sound) ;)

Ok let me on a better way.

I need to shoot a commercial pub and i want to show more about the possibilities of the GH2 That i check what is the resolution limitation on record mode.

I first test 4K but it fail, when i press record nothing happen, well i decided to test 3K resolution it reccord but stop... after all i decided to test the BMCC resolution 2.5K, it work perfectly

I forget to say that i use settings from Apocalypse 444 with some personnal modification.

When is say RAW i mean the Raw of the GH2 not the Raw Raw ;)

fatpig i really think that the GH2 record in 2.5K i don't have that much details in 1080p/720p with the same settings and more if i downscale to 1080p i have a more detailled 1080p than what the GH2 usually record in 1080p

perder

pics or it didn't happen..

inqb8tr

@Neverprod You shoot in mjpeg mode, that you have set to 2.5K in ptools? It is interesting to see how far can it go. Generally AVCHD on GH2 gives better IQ than MJPEG does, especially in low light, shadow areas etc. Maybe some comparison would make sence, 1080p avchd vs 2.5K MJPEG downscaled to 1080.

Neverprod

Perder i already show the proof before... i deleted the video because i keep the sound that i was not supposed to leave it ;) (private video)

Here GSpot print screen

Neverprod

@inqb8tr

Yes a shoot in MJPEG Mode (one of my favorite for exploit the GH2 and other cam)

I set 2.5K in Ptools and i'm checking at this moment what is the limit but i feel that i can't go up than 3K "Generally AVCHD on GH2 gives better IQ than MJPEG does" -> not sure of that when you know that AVCHD don't take all the picture details.

Low Light and Shadow Areas is just a question of Postprod, i prefer to use MJPEG mode than AVCHD in postprod.

Tomorrow i will take time to show the comparison between the 1080p AVCHD and the 2.5K MJpeg (with downscale to 1080p) too bad you didn't see the video i upload before that show how nice is the 2.5K MJPEG

Meierhans

While I doubt very strong that the upscaled MJPG will exeed AVCHD, there has been alot tests about a year ago, but this might indeed be useful when it comes to chroma keying. If the scaling happens before the downsampling to 4:2:0 it should preserve some more color values.

Neverprod

It's not what Apocalypse do already ? matrix 444 + 2.5K MJPEG = nice couple

Neverprod

Did we find a possibility to add a new codec mode on the GH2 or not yet ?... Change the codec algorithm and bring a new one that exploit more the GH2 ? i have impression that the GH2 have not limit, feel weird.

TommiH

Obvious troll is obvious....

Neverprod

Second result about a better resolution I try different know and unknown resolution that is related to the 2:39/1:78 etc... For now i find that the best height for the GH2 is 1366 after that, i can see some problem (maybe can be fixed) About the width i'm not sure yet but from my investigation, i find that the GH2 accept easily the 2560 size (2560x1366 with a display ratio of 1:85)

Fast test of 10seec

Video ID : 1 Format : JPEG Codec ID : jpeg Duration : 9s 500ms Bit rate mode : Variable Bit rate : 119 Mbps Width : 2 560 pixels Height : 1 366 pixels Display aspect ratio : 1.85:1 Frame rate mode : Constant Frame rate : 30.000 fps Compression mode : Lossy Bits/(Pixel*Frame) : 1.130 Stream size : 134 MiB (94%)

@TommiH

Maybe you would need to calm down ? if you are not interested for talking about, you can go play your troll somewhere else ?

TommiH

@Neverprod I'm perfectly calm. :) You just registered to this site, and yet you came up with a way to get the resolution bumped up to 2.5K? The image is upscaled, thats all. Your claim of real 2.5K resolution is nonsense.

If I understood you incorrectly, and you are talking about the advantages of upscaling the uncompressed image and then resizing it to the original source size, I apologize, but real 2.5K is not accessible with the tools we have available to us. :)

Here is a 2K test of mine:

;) Peace! :)

fatpig

Just post a screenshot from some detailed scenery/leaves shot in your 2.5k mjpeg vs shot on regular avchd in 24H - then we can see what you mean.

@all: there is a 2k mode in driftwoods patches. does anyone know which benefits it is supposed to have?

Neverprod

@TommiH

You still think that in MJPEG Mode 2.5K the image is upscaled... You understand that every DSLR/Hybrid and co have a 4K Sensor already ? the recording solution depend of the software and hardware using.

This is a true capture of 2560x1366 and not a upscale...i'm not talking about the AVCHD mode

Yes i'm a registered on the website so ?... i'm working for over in over years in the cinema industry, i know exactly what i'm talking about... also don't judge a person too fast ok... you don't even know me, thanks.

I'm talking about taking the advantage of the sensor capabilities not the upscaling..

It's you saying no sense... you don't even try to understand and just came to critic, you don't even know what i'm showing now.

Also you seem to don't understand what the resolution mean and what is for on a Sensor

Neverprod

@TommiH

Now you still think that is a upscaling, seriously don't think i'm dumb thanks i know exactly what i'm saying

Again be more respectfull to other next time and not only me

TommiH

@Neverprod I'm not trying to be rude, I just don't understand what you are after...

Do you mean you actually think that the video pipeline is actually delivering you 2.5K resolution? I mean Full sensor readout at 30fps and scaling down to 2.5K? The AVCHD version is clearly showing less detail, but that is because of compression, not because of actual resolution difference.

Neverprod

@TommiH

Did you really look both picture scene ?

Clearly not a upscale, if it was a upscale of the 720P/1080p i will not catch more stuff on my scene (look who many stuff you can see on 2.5K and on the 1080p) and it's exactly the same recording position moment ! in here we clearly (repeat again) use the sensor possibilities to take all the scene and not a simple software upscaler.

If i will can use the limit of the sensor, i will can take more stuff on my scene than the 2.5K

It's actually really giving me a true 2.5K resolution even a little bit more than the BMCC resolution.

That's why is was impressed to see a 2.5K at 30p working stable on the GH2, i didn't expected that much from the hardware of the GH2.

Now if i try to reach the 3K , i have too much instability that maybe can be fixed

TommiH

@Neverprod What I'm seeing is that the aspect ratio on the 2.5K sample is squeezed and it's shot in a different angle in relation to the subject. Also I think the aperture is different... To my eyes it's hardly scientifically comparable. In my opinion, what you think is more resolution is simply less compression and faulty method of metering a difference of settings.

Mount your camera on a tripod, lock all variables on the camera, shoot some fine cloth or other repetitive matrix and post the results. I actually want to be proven wrong, because I'd rather like 2.5K resolution... :)

Please don't get offended, I'm just quite sure I'm right. Prove me wrong. :)

Stray

We've been through this with mjpeg allowing arbitrary resolutions, and IIRC noone was really convinced that it was actually using the resolution of the sensor as per the settings you gave. It seemed much more likely that the camera was performing scaling to the requested resolution. Also upscaling AVCHD gave better results than getting the camera to record mjpeg at a 2K resolution is my memory of this debacle.

Anyway, that screenshot comparison of your mouse on the table with an out of focus background is completely useless for any kind of good comparison (except maybe a lowlight performance comparison, which again there has been much talk/comparison here between mjpeg and AVCHD too) go out in the daylight and shoot something with a lot of detail, an awful lot (like trees, or bushes) with the same lens in AVCHD 1080p and then with your 2.5K setting. Then it will be easier to judge, and would be worth doing, cos my memory of this discussion on mjpeg and its resolution possibilites is that it remained unresolved. And as @TommiH said, the aspect ratio needs fixing. One thing I know people do do with the mjpeg settings is use them to shoot with anamorphic lenses so as to get the right resolution/ratio needed to unsqueeze the resulting footage.

Neverprod

@TommiH

In France is late right now both 1080p/2.5K using same lens olympus 45mm 1.8

Edit : i do that tomorrow, it's really late, need to sleep...

Again is the same position, same moment check correctly, you go understand why is look like moving a little when is not but like i said again i go show more testing 2.5K MJPEG

@Stray

I was already convinced that we really use the sensor capabilities in here but i didn't expected a good stability Yes the picture in here is useless but show that you can clearly see the difference if you take a look you don't need a zoom but check

The 2.5K is more wide you can see more stuff on the scene (check the border etc...) The 1080p is clearly no wide as the 2.5K and the stuff i see on 2.5K is out of scene in here

Howdy, Stranger!

Categories

Tags in Topic

Top Posters