Does the GH2 exhibit any aliasing?

Of course GH and AF show aliasing. Relatively big sensor & high MP = aliasing. Not as terrible as APS-C and FF sensors. M43 is just right size. M43 lenses tend to be super sharp, and some filters can mitigate the issue with added benefits.

 
Among all filter types, ND filter is so critical to video that I wish Panasonic makes built-in ND filters in GH3. Then adding more filters would become much easier. The post processing can't mimic what those filters can do.
 
Of course GH and AF show aliasing. Relatively big sensor & high MP = aliasing.

I've heard others say this, what does it mean? Seems like you are missing a step. You're talking about how higher pixel count without higher pixel rates necessitates more line skipping, right? Outside of line skipping, if anything, larger sensor size or higher pixel count could decrease aliasing, all else being equal, right?
 
High MP means it has many holes to fill in. For highly detailed scene, digital gadget tries guesswork to fill in the holes. Sometimes the guesswork generates ugly aliasing. Sometimes due to mechanical shutter. Some filters can blur the projected images to mitigate the issue, but not completely removing it. Complete removal of aliasing means too much loss in details.
 
High MP means it has many holes to fill in. For highly detailed scene, digital gadget tries guesswork to fill in the holes. Sometimes the guesswork generates ugly aliasing. Sometimes due to mechanical shutter. Some filters can blur the projected images to mitigate the issue, but not completely removing it. Complete removal of aliasing means too much loss in details.

It has NOTHING to do with large MP AT ALL. Look at the results of the Red sensor in terms of aliasing. If the GH2 could read out all pixels and downsample to 1080p the image would be absolutely amazing. The problem comes from the methods of pixel binning/reduction of sample information AND the algorithms for building the final image ESPECIALLY with a Bayer pattern. In fact, a larger MP imager that USES all those pixels and then downsamples to the target of say 1080p is going to be SUPERIOR in terms of sharpness to something that samples at the target if they both use the same filter technology. Now, that doesn't mean sensitivity and noise rejection are better, but aliasing will be GREATLY reduced.

The real issue is that with sensors that use the bayer pattern, they pick a OLPF frequency between green and blue/red spatial frequency so they don't make the image too soft. I've seen blue/red chroma aliasing on red images at FULL resolution before.

I have yet to EVERY see a Canon outperform the GH1/3 or GH2 on the aliasing front. I've shot with my 7D since they came out and KNOW the aliasing in the camera hands down, which is equivalent to the 550D. The biggest issue with the Canon is the chroma aliasing and if you look closely, you'll be amazed at how many shots have it. It's an absolute pain in the neck with VFX work sometimes. (And the 4:2:0 just compounds on that situation).

I shot the tests above with my AF100 and I was pretty impressed with the aliasing handling on the GH2. I'm also VERY happy with the handling on the AF100 as well. My 7D would have failed horribly trying to shoot the brick AND the rooftops.
 
Red one has only 11.5 MP.

All condition being equal, lower MP should make it easier to deal with the aliasing issue. High MP does make it more difficult. e.g. AA filter selection.
 
Have to differ the moire aliasing looks worse on my GH2 than the 550D/T2i i had before, i like nearly everything else better though on the Pana.

I have no idea what footage you've experienced but the aliasing on the GH1 and GH2 is nothing like the extreme aliasing on the Canons.

I have both GH's, my shooting partner has a 60D and its aliasing is about 10x worse. I have an unscientific, but real life example here.
 
Why we have 18MP? Is it because of 3.9x ETC?

2MP sampling * 3.9 * 3.9 = 16MP in 16:9 mode. It's 18MP in 4:3.

If yes, screw 3.9x ETC. 14MP is good enough for 2x ETC.

3MP sampling * 2 * 2 = 12 MP in 16:9. That's roughly 14MP in 4:3. Panny, please let's go back to 14MP and stay in 14MP. Just like M43 size compromise, MP size should be compromised. I like 14MP.
 
Why we have 18MP? Is it because of 3.9x ETC?

2MP sampling * 3.9 * 3.9 = 16MP in 16:9 mode. It's 18MP in 4:3.

If yes, screw 3.9x ETC. 14MP is good enough for 2x ETC.

3MP sampling * 2 * 2 = 12 MP in 16:9. That's roughly 14MP in 4:3. Panny, please let's go back to 14MP and stay in 14MP. Just like M43 size compromise, MP size should be compromised. I like 14MP.

Yes that is precisely why the GH2 has 16MP and I for one love the 3.9x ETC. I shot some tests comparing the GH2 ETC aliasing to the GH2 without ETC. As expected the ETC is nearly perfect. Here are those samples.

http://vimeo.com/20183170
 
Last edited:
The ultimate aliasing / moire test shot I have done on my GH13 is my laundry scene which starts at about 1:20 an goes to 1:40

I pull focus to every horrible pattern, horizontal line, vertical line, tiny detailed patterns and lines (I hung those ugly clothes up on purpose) and the GH13 passes with flying flying colors. Yeah there is a tiny tiny bit of aliasing but almost none, only the most HARD core of pixel peepers would spot it and they would have to have their face up to the monitor.

I agree that clip is definitely a torture test and the GH13 does extremely well. By the way that clips looks great on Vimeo. I never thought it was anything special on youtube. It looks great on Vimeo though. I will have to download the original some time.
 
Red one has only 11.5 MP.

All condition being equal, lower MP should make it easier to deal with the aliasing issue. High MP does make it more difficult. e.g. AA filter selection.

High MP does NOT make it more difficult. The problem is, we aren't reading out all the MP as we would a photo off of the DSLR cameras. THAT is the issue. In fact, the higher your MP, the more sampling you are doing color wise per unit area, which does in fact reduce aliasing. If you then take that high MP image and reduce it down to a lower MP image, it will look SUPERIOR to an image shot at the target resolution in terms of aliasing, period, both cameras using the same filter array mechanism.
 
ETC seems limited to low iso in practice. Also 2.6x for 1080p and 3.9x for 720p only.

14MP sensor for better DR and less aliasing. 2x ETC in 1080p and 720p. 3MP sampling will make it useful in higher iso. But it seems Panny is heading toward higher MP.
 
High MP does NOT make it more difficult. The problem is, we aren't reading out all the MP as we would a photo off of the DSLR cameras. THAT is the issue. In fact, the higher your MP, the more sampling you are doing color wise per unit area, which does in fact reduce aliasing. If you then take that high MP image and reduce it down to a lower MP image, it will look SUPERIOR to an image shot at the target resolution in terms of aliasing, period, both cameras using the same filter array mechanism.
I never heard that before. Frankly I don't buy into it.
 
I never heard that before. Frankly I don't buy into it.

Well frankly, then you don't understand image sensors then. A higher MP camera has higher spatial resolution which is SUPERIOR to getting rid of aliasing. It's that simple. The problem with DSLR cameras is they don't actually sample normally all the pixels, they use fast less than desirables techniques like pixel binning and the like to avoid reading ever pixel on every row since the system as a whole is not capable of handling the throughput. Take a photo with a DSLR and then you see the full resolving power of the MP of the camera. It looks beautiful. Once the DSLR gets to the point it can take video at the same resolution as the stills, and either use the full resolution or use a GOOD post sensor reduction algorithm, the image is going to be fantastic looking.

If you don't believe it, take a 60D, or GH2, take a short bit of video, then take a picture, shrink that picture down to 1080p and be prepared to see how much better the image looks than the video grab. THAT is the problem.
 
I agree that clip is definitely a torture test and the GH13 does extremely well. By the way that clips looks great on Vimeo. I never thought it was anything special on youtube. It looks great on Vimeo though. I will have to download the original some time.

DO IT. DO IT.

I still stand by my stance that it is the clearest most detailed filled and artifact free encode on youtube. That's not saying it doesn't have some 'stuff'' going on....
but it's the current BEST that youtube can do with 1080p until they up the bitrate a bit.

Definitely the vimeo is better, but vimeo 1080p is harder to play on machines for whatever reason.
 
Well frankly, then you don't understand image sensors then. A higher MP camera has higher spatial resolution which is SUPERIOR to getting rid of aliasing. It's that simple. The problem with DSLR cameras is they don't actually sample normally all the pixels, they use fast less than desirables techniques like pixel binning and the like to avoid reading ever pixel on every row since the system as a whole is not capable of handling the throughput. Take a photo with a DSLR and then you see the full resolving power of the MP of the camera. It looks beautiful. Once the DSLR gets to the point it can take video at the same resolution as the stills, and either use the full resolution or use a GOOD post sensor reduction algorithm, the image is going to be fantastic looking.

If you don't believe it, take a 60D, or GH2, take a short bit of video, then take a picture, shrink that picture down to 1080p and be prepared to see how much better the image looks than the video grab. THAT is the problem.

Yes this is clearly true, but I think what he was saying is that he doesn't agree that this holds true once you factor in the downscaling (and I don't think you think it does either.) Higher megapixel sensors mean more downscaling that has to be done, and when done poorly, could lead to more aliasing. Canons use line-skipping vs GH pixel binning and have much worse spatial resolution (in video.) The higher resolution the sensor is the more data is going to have to be thrown out to reach 1080p in the same amount of time. With the same processing power this could lead to a degraded video image.
 
Well frankly, then you don't understand image sensors then. A higher MP camera has higher spatial resolution which is SUPERIOR to getting rid of aliasing. It's that simple. The problem with DSLR cameras is they don't actually sample normally all the pixels, they use fast less than desirables techniques like pixel binning and the like to avoid reading ever pixel on every row since the system as a whole is not capable of handling the throughput. Take a photo with a DSLR and then you see the full resolving power of the MP of the camera. It looks beautiful. Once the DSLR gets to the point it can take video at the same resolution as the stills, and either use the full resolution or use a GOOD post sensor reduction algorithm, the image is going to be fantastic looking.

If you don't believe it, take a 60D, or GH2, take a short bit of video, then take a picture, shrink that picture down to 1080p and be prepared to see how much better the image looks than the video grab. THAT is the problem.

Grimeppoch, soon this will be fixed.... with words like 'throughput' limitations being ancient terminology like typing

load "*",8,1

on a Commodore 64.

The future comes fast, and I say, bringeth it, but bringeth it NOW!

AYE!
 
Yes this is clearly true, but I think what he was saying is that he doesn't agree that this holds true once you factor in the downscaling (and I don't think you think it does either.) Higher megapixel sensors mean more downscaling that has to be done, and when done poorly, could lead to more aliasing. Canons use line-skipping vs GH pixel binning and have much worse spatial resolution (in video.) The higher resolution the sensor is the more data is going to have to be thrown out to reach 1080p in the same amount of time. With the same processing power this could lead to a degraded video image.

True. One has to take into consideration though that a DSLR is a STILLS camera first. Being that as it is, it makes even more sense to have higher MP for higher quality stills. Ultimately I want both higher resolution AND higher throughput, but, we can clearly see that is not possible yet. It's going to be great when it does though!

:)
 
Grimeppoch, soon this will be fixed.... with words like 'throughput' limitations being ancient terminology like typing

load "*",8,1

on a Commodore 64.

The future comes fast, and I say, bringeth it, but bringeth it NOW!

AYE!

Hehe, funny thing, I was using a C64 emulator the other day trying to remember how to do something :) I could not remember for the life of me (although I was more of an Atari 130XE guy) :)
 
Back
Top