GH3 vs hacked GH2 quick handheld comparison

[see ya]

@MaxAperture, steady on, Karim has already said he'll post the native camera files, I too have no interest discussing 'tests', so many variables in every respect can skew, but he's prepared to spend time shooting and providing, lets not bite the hand that feeds.

Do we need images repeatedly requoted from a post or two above, really.

[KarimNassar]

[quote name='MaxAperture Films' timestamp='1351265274' post='20374']


You cropped the footage and used image stabilization, this changes the pixel scaling when it is rendered back out in the final comp and causes aliasing artifacts. This was meant as a response to jgharding's freak out, and is what you see on the rims of the glasses in both camera's footage.

To be honest, I'm getting a bit tired of watching poorly implemented comparisons that end up promoting myths about this camera's image quality. If you don't have the time to do a proper comparison with a tripod and have to rely on warp stabilization just to compare the shots, you're really not doing anyone a service because the footage is not representative of true pixel-pixel output quality. It's just muddys the water further and leads to more confusion.
[/quote]

No...............

This would be very true if relevant but it is not.

I did not use wrap stabilization.
Only tracked the footage and moved it along the x and y axis which does not deform pixels.

Also I did not crop the footage but only masked the area I did not want.

And finally the raw footage straight out the camera is available for download so there is no room for misinformation.

[KarimNassar]

[quote name='yellow' timestamp='1351265802' post='20376']
@MaxAperture, steady on, Karim has already said he'll post the native camera files, I too have no interest discussing 'tests', so many variables in every respect can skew, but he's prepared to spend time shooting and providing, lets not bite the hand that feeds.

Do we need images repeatedly requoted from a post or two above, really.
[/quote]

Thank you and in case you missed it the link has been provided in the original post

He has very true good points but they just don't apply with what I provided

[MaxAperture Films]

[quote name='KarimNassar' timestamp='1351266065' post='20377']
No...............

This would be very true if relevant but it is not.

I did not use wrap stabilization.
Only tracked the footage and moved it along the x and y axis which does not deform pixels.

Also I did not crop the footage but only masked the area I did not want.

And finally the raw footage straight out the camera is available for download so there is no room for misinformation.
[/quote]

Thanks for clarifying Karim. I still think something is somehow off due to the fact I have never seen my GH2 alias the way the hacked GH2 and GH3 look in your comparison. This may be somehow due to how Media Player is decoding the F4V files (which I never use). Tough to say for sure... and thanks for all your efforts, please don't take my concern as an attack upon you in any way.

[KarimNassar]

[quote name='MaxAperture Films' timestamp='1351265671' post='20375']


I would recommend trying a sharpness test with both cameras set at 0, rather than -2 and -5. We don't know if the GH2 and GH3 are artificially softening the image at the lowest value or just removing all in-camera sharpening altogether. Comparing both at 0 should increase the liklihood of a fair comparison (no softening or sharpening) assuming Pansonics implementation methodology has not changed.
[/quote]

Interesting approach and worth trying out.

I am the first to say that it was a limited test and the best I could do with the constraints that I had.
I clearly stated that.
So not necessarily the most flattering or revealing nor accurate test.

What's the point then you may rightfully ask?

Well for instance, if the gh3 really had a new high dynamic range sensor, that would have been clearly visible even when shooting handheld in a rush.
But it is not.
So it is definitively revealing and useful in that aspect for example.

[MaxAperture Films]

[quote name='yellow' timestamp='1351265802' post='20376']
@MaxAperture, steady on, Karim has already said he'll post the native camera files, I too have no interest discussing 'tests', so many variables in every respect can skew, but he's prepared to spend time shooting and providing, lets not bite the hand that feeds.

Do we need images repeatedly requoted from a post or two above, really.
[/quote]

Good point, pics now deleted from post response.

[KarimNassar]

[quote name='MaxAperture Films' timestamp='1351266615' post='20379']


Thanks for clarifying Karim. I still think something is somehow off due to the fact I have never seen my GH2 alias the way the hacked GH2 and GH3 look in your comparison. This may be somehow due to how Media Player is decoding the F4V files (which I never use). Tough to say for sure... and thanks for all your efforts, please don't take my concern as an attack upon you in any way.
[/quote]

Your critique was very welcome.
And most importantly absolutely true, just had false assumptions on me but they are justified based on what content we usually get online.
Most people on this boards are way more knowledgeable than I am and I am here to learn and share.
I have no business being here if your intelligent argument was an issue for me.
So no worries and quite the contrary thanks for contributing

[MaxAperture Films]

[quote name='KarimNassar' timestamp='1351266692' post='20380']
Interesting approach and worth trying out.

I am the first to say that it was a limited test and the best I could do with the constraints that I had.
I clearly stated that.
So not necessarily the most flattering or revealing nor accurate test.

What's the point then you may rightfully ask?

Well for instance, if the gh3 really had a new high dynamic range sensor, that would have been clearly visible even when shooting handheld in a rush.
But it is not.
So it is definitively revealing and useful in that aspect for example.
[/quote]

I don't think we can conclude that until all picture profiles are explored.

For example, you used the widest dynamic range picture profile on the GH2 (nostalgic) but only set a custom profile on the GH3 (picture profile unkown). I am sure there are some picture profiles on the GH3 that have a wider dynamic range than others, but this will all become evident with more testing.

On the dynamic range test you posted it also appears to me that the back wall lights are more blown out on the GH2 and not on the GH3, while the GH2 ceiling lights also appear more blown out in comparison.

[KarimNassar]

[quote name='MaxAperture Films' timestamp='1351267174' post='20383']
I don't think we can conclude that until all picture profiles are explored.
[/quote]

True.

I did swap through them and felt like the custom one with contrast set to the minimum should be the best in that aspect but too soon to say indeed.

[jgharding]

[quote name='MaxAperture Films' timestamp='1351265274' post='20374']
You cropped the footage and used image stabilization, this changes the pixel scaling when it is rendered back out in the final comp and causes aliasing artifacts. This was meant as a response to jgharding's freak out, and is what you see on the rims of the glasses in both camera's footage.
[/quote]

after watching plenty of tests, and that high profile Genesis or whatever it was called, I'm just not impressed... Stabilising [i]is [/i]common practice in real world situations, so if the GH3 looks that much worse than the GH2 when both have been stabilised, well then it looks worse in a real-world situation.

I regularly use software scaling, whichever way you look at it, that level of aliasing is bloody awful, and a decent source image (like C300) won't suddenly alias heavily under scaling. The beard is gibbering with jaggies!

This kind of test is [i]much [/i]more useful to me than a chart or two tripod shots of waving leaves. Personally, I'm not too fussed about theoretical or "pixel for pixel output quality" cos it's not how I use the cameras. I do however, shoot a lot of handheld shots of people that i then stabilise them, and here the GH3 was kicked soundly by it's predecessor.

*

On the dynamic range shot the GH3 has cleaner shadows and a greater number of usable stops due to less banding by the looks of it, so it won there. Seems to have the same overall stop range, but smoother transitions within it. But it's a softer, mushier image overall. I'm just not feeling it.

I'm sure we could all make a good piece with it in a great controlled environment, granted. But the fact remains: on the evidence so far, the GH3 has more usable dynamic range (cleaner shadows) and less banding, but a poorer image overall. when it comes to detail and certainly aliasing, it loses to its predecessor. I just think it's a shame. I'm not tempted to try it really.

I'm quite a nuetral party here, I use hacked 550D, C300, FS700 and RX100 more than anything else, so no vested interest. I don't [i]need[/i] the GH3 to be good in order to sleep at night, for example. I'm just saying what i see...

[jgharding]

We can only judge from what we see though, perhaps popping the raw files on a server for free would be a good plan, then we can eliminate any processing in between as a factor?

[Ernesto Mantaras]

Why is there such a different exposure in each moustache man shot? Weren't the cameras at the same aperture, ISO and shutter settings?

Also, I notice something I didn't expect: I see aliasing in the upper frame of the glasses in the GH2 shot, although moiré is nowhere to be seen. Yet I see moiré in the GH3 shot but no aliasing... :S
So now one is the other one's tradeoff? Damn!

By the way, I watched the raw files in the After Effects timeline at the best possible quality and zoomed when needed.

[see ya]

First thoughts from looking at the native files subjectively the GH3 h264 MOV appears to have slightly more detail, the GH2 MTS AVCHD has plastic looking skin and 'smudged' areas through out.

The GH3 h264 MOV appears to have no 16x16, 8x8, 4x4 partitioned macroblocks only 8x8 and they are prominent in a still image, but this maybe something odd with my set up although I've not seen this with any other source, only GH3 h264.

The GH3 h264 MOV as others have mentioned has a lot of moire, the GH2 AVCHD non discernable, would have been interesting to compare GH3 AVCHD MTS too.

GH3 h264 MOV makes use of full 8bit range, the GH2 MTS the limited 8bit range. So should be finer gradation in the GH3 h264 MOV, certainly the highlights are spread over more levels, the GH2 MTS highlights are compressed excessively but then it's brighter exposure not a like for like comparison.

[galenb]

[quote name='MaxAperture Films' timestamp='1351265671' post='20375']
I would recommend trying a sharpness test with both cameras set at 0, rather than -2 and -5. We don't know if the GH2 and GH3 are artificially softening the image at the lowest value or just removing all in-camera sharpening altogether. Comparing both at 0 should increase the liklihood of a fair comparison (no softening or sharpening) assuming Panasonics implementation methodology has not changed.
[/quote]

I was just about to mention this too. with sharpness set to -5 it actually softens the image. This is what Julian found out with his tests. At least, that's the way it appeared.

[galenb]

[quote name='MaxAperture Films' timestamp='1351265274' post='20374']
You cropped the footage and used image stabilization, this changes the pixel scaling when it is rendered back out in the final comp and causes aliasing artifacts.
[/quote]

Actually I don't think this is entirely true. Scaling, if done correctly in software by either summing or interpolating, and not simply pixel skipping, should not cause aliasing in the image. Especially not in only specific areas of the image. If done wrong (like in mspaint or some really old juvenile painting program), it would be visible across the entire image. I do a lot of VFX and post work and have, to the best of my knowledge, never seen aliasing caused by scaling the image in post programs like After Effects, Premiere, Nuke or Final Cut. This is the 21 century and most software developers are aware of how to properly scale images. Also, if stabilization caused aliasing artifacts then you would see it all time considering how common it is. If anything scaling of the image causes softening.

It is however totally common to see aliasing happen in the encoding/decoding/transcoding process. In fact, I'd say this is probably the most common way to introduce aliasing artifacts into the footage.

Now, this is not to say that scaling or stabilization doesn't cause muddying of the original image quality and as you say, should be totally avoided. Just that it would not be the cause of this kind of aliasing in this case.

[sandro]

I'm afraid you are right. After watching the bearded man comparison the GH3 aliasing is just horrible! What happened to Panasonic???

[Germy1979]

I've said this before, but why would any camera manufacturer... (Especially one that makes a camera loaded with prioritization in its video features) - NOT put aliasing and moire as part of a basic checklist? Who in the hell wants aliasing and moire? Lol...

Man i hope they fix this...

[sanveer]

I have a strong gut feeling, that the image and video quality, is more than just a codec issue. It may have to do with the sensor, as well as the wiring (circuitry). I guess, we'll know by the start of December, when (hopefully), the Firmware is finalised, for release.

[Ernesto Mantaras]

This moiré problem we're seeing is definitely a processing problem in the hardware department. But I bet it is plenty capable of resolving images without moiré, so I guess the scaling method could be improved or the architecture modified to solve this without losing all of the other things the GH3 does get right.
It does bother me that the so called "wide dynamic range" isn't really there either (although I saw picture shot with the High Dynamic Range profile and it looked Cinestyle like a bit).

Could it be that the small delay has been stated to solve it?

[sanveer]

[quote name='charlie_orozco' timestamp='1351365867' post='20443']
This moiré problem we're seeing is definitely a processing problem in the hardware department. But I bet it is plenty capable of resolving images without moiré, so I guess the scaling method could be improved or the architecture modified to solve this without losing all of the other things the GH3 does get right.
It does bother me that the so called "wide dynamic range" isn't really there either (although I saw picture shot with the High Dynamic Range profile and it looked Cinestyle like a bit).

Could it be that the small delay has been stated to solve it?
[/quote]In the GH2, only a portion of the sensor was used for the video, and the down scaling wasn't the usual method, used in other sensors. Therefore, presumably, the circuitry was in order.

Maybe, the GH3 has a better sensor, than the GH2, notwithstanding the smaller size (and disengaging the multi-aspect ration bit, and disabling the ETC, for video). But, merely because a sensor (and even the processor), is better, doesn't mean that the camera's video and stills would be. In the newer sensors, circuity is extremely complicated (especially in the back-lilluminated ones). The GH3 may not have a back-illuminated one (GOD only knows the exact kind of sensor it has), but,I am guessing, that Panasonic has to seriously tweak a lot of things, including circuitry and codec.

After the benchmark the GH2, set, for video, the only sensible way, would have been, either a 2.5-2.7k video,or a 2k video, with the dynamic range of atleast 2-3 points higher, and much flatter colour profile. This without Any Moire or Aliasing. This is something, which is Clearly lacking in the GH3.

I am guessing (that) the delay, was due to not getting their act in order. They chose Photokina, because, otherwise, they would have had to showcase at some smaller, less prestigious, venue. If Panasonic is afraid that the GH3 will eat into their Professional Video Camera market, then, they should think of the revenues only in terms of numbers and not the profit margins. If the GH3 shoots anywhere near the Alexa or C300 (or even the C100), then, it would easily outsell the GH2, Many Times Over, and most people who picked up one GH2, would easily be tempted to get 2 GH3s, instead. This apart from all the new buyers. Also,people thinking of picking up any of the Canon duds including the 5D Mark II and Mark II and their C100 and C300, would have a good enough reason to move to the GH3. Maybe other models like by Nikon and Sony also could be passed over, for a GH3. Panasonic should think of things, from a broader perspective, and from a quantity perspective, rather than larger margins, per camera.