550D/GH13 comparision shot

[Grunf]

Actually, it is a bit of a different comparison. The problem with the Canons is the rainbows. When I try and pull that out of the equation, the moire is definitely a lot more pleasing, and a lot less noticeable. However, the rainbow kicks in very early and is horrible looking. If someone wants to hack the Canon, that is the first thing I'd be trying to figure out is how to get rid of that!

It all boils down to Nyqvist limit when it comes down to sampling pixels. Even if you had perfectly quadratic RGB pixels (a'la Foveon), you have to have some sort of low-pass filter to get rid of aliasing. Bayer-pattern makes this worse. Sub-sampling Bayern-pattern sensor by line-skipping makes it even more so. That's why we get rainbows.

Panasonic seems to have chosen the right path: scan the whole 12MP sensor, De-Bayer, rescale with averaging and feed it to codec. Unfortunately, they seem to have added extra sharpening between the rescale and codec. Thus the jagged lines despite oversampling. Fortunately, it can be toned down by pulling down the sharpness to -2. (Edit: just saw some trumpet tests with GH1. It seems to be suffering from false detail as well, so it probably has some dodgy scheme of reading pixels. Just better than Canon)

Canon, on the other hand, seem to have line skipped w/o any sort of rescale, and then fed raw that to codec. As RGB-pixels aren't perfectly in-line (It's Bayer sensor, after all), you get dancing colours. It's a pity as sensor is even more capable than GH1 one. If they could read the whole sensor, then do De-Bayering and resampling down to 1080p then it would be digital Nirvana. I wonder if CPU has horsepower to do so. Even if it has, it would be major re-write of firmware.

 

[Rick Burnett]

It all boils down to Nyqvist limit when it comes down to sampling pixels. Even if you had perfectly quadratic RGB pixels (a'la Foveon), you have to have some sort of low-pass filter to get rid of aliasing. Bayer-pattern makes this worse. Sub-sampling Bayern-pattern sensor by line-skipping makes it even more so. That's why we get rainbows.

Panasonic chose the right path: scan the whole 12MP sensor, De-Bayer, rescale with averaging and feed it to codec. Unfortunately, they seem to have added extra sharpening between the rescale and codec. Thus the jagged lines despite oversampling. Fortunately, it can be toned down by pulling down the sharpness to -2. As

Canon, on the other hand, seem to have line skipped w/o any sort of rescale, and then fed raw that to codec. As RGB-pixels aren't perfectly in-line (It's Bayer sensor, after all), you get dancing colours. It's a pity as sensor is even more capable than GH1 one. If they could read the whole sensor, then do De-Bayering and resampling down to 1080p then it would be digital Nirvana. I wonder if CPU has horsepower to do so. Even if it has, it would be major re-write of firmware.

Canon has said they do not line skip and use pixel binning and, as has been shown, with certain settings with the hack the GH13 can be made to have rainbows as well, which was very interesting to say the least.

I definitely understand spacial frequency and the resolution capability pertaining to spacial sampling, as it not only pertains to pixel pitch but also pixel size as well. In addition, I understand the limitations by using a bayer pattern that essentially has 2 different pixel resolutions, one for the green and another for both blue and red. It seems most manufacturers of sensors place the OLPF somewhere inbetween as to maximize the results and not create an overly soft image just to minimize all aliasing.

I've not seen definitive proof of any of the claims pertaining to what Canon is actually doing. Until someone can hack the software, we are just not sure. They claim (as Barry has mentioned in another thread) they use pixel binning. Given what I have seen the GH1 do I am more inclined to believe that maybe they are using a poor pixel binning methodology or something is up with their flavor of H.264.

 

[Barry_Green]

Will go out and get some side-by-side shots on GH13 and 7D. Ducks and trees (sorry Isaac). But these will be using the two best still lenses I have, the hyper-sharp Zeiss ZF 100mm (on the 7D) and 85mm (on the GH13). Both will be at f/8 and properly focused, so any differences will absolutely not be attributable to the lenses.

Neutral on the 7D, with sharpness set to the middle. Smooth on the GH13, with sharpness set in the middle.

Whatever happens, happens. I'm just gonna go get the shots and post 'em, and y'all can decide what it "means."

 

[Rick Burnett]

Will go out and get some side-by-side shots on GH13 and 7D. Ducks and trees (sorry Isaac). But these will be using the two best still lenses I have, the hyper-sharp Zeiss ZF 100mm (on the 7D) and 85mm (on the GH13). Both will be at f/8 and properly focused, so any differences will absolutely not be attributable to the lenses.

Neutral on the 7D, with sharpness set to the middle. Smooth on the GH13, with sharpness set in the middle.

Whatever happens, happens. I'm just gonna go get the shots and post 'em, and y'all can decide what it "means."

Not sure how much time you have, but setting the sharpness to both extremes on both might be nice as well just to see what it does to the results.

 

[Kholi]

Not sure how much time you have, but setting the sharpness to both extremes on both might be nice as well just to see what it does to the results.

Yeah I'd love to see how nasty both images are with sharpness at max. Where's Ian at?

Hehehe

 

[Grunf]

Will go out and get some side-by-side shots on GH13 and 7D. Ducks and trees (sorry Isaac). But these will be using the two best still lenses I have, the hyper-sharp Zeiss ZF 100mm (on the 7D) and 85mm (on the GH13). Both will be at f/8 and properly focused, so any differences will absolutely not be attributable to the lenses.

Neutral on the 7D, with sharpness set to the middle. Smooth on the GH13, with sharpness set in the middle.

Whatever happens, happens. I'm just gonna go get the shots and post 'em, and y'all can decide what it "means."

I just reviewed some footage (cable lines etc.) I took today and came to conclusion that Panny uses too much sharpening even when sharpness is at "0"...probably just to please consumer crowd. I believe the correct image on Panny should use sharpness at "-2" and noise reduction at "-2" (unless we are talking low-light). But it'll certainly be very interesting to see your test footage!

 

[Isaac_Brody]

Will go out and get some side-by-side shots on GH13 and 7D. Ducks and trees (sorry Isaac). But these will be using the two best still lenses I have, the hyper-sharp Zeiss ZF 100mm (on the 7D) and 85mm (on the GH13). Both will be at f/8 and properly focused, so any differences will absolutely not be attributable to the lenses.

Neutral on the 7D, with sharpness set to the middle. Smooth on the GH13, with sharpness set in the middle.

Whatever happens, happens. I'm just gonna go get the shots and post 'em, and y'all can decide what it "means."

Barry, could be good to also get footage at the same time with sharpness dialed all the way back just for comparison.

 

[Grunf]

Will go out and get some side-by-side shots on GH13 and 7D. Ducks and trees (sorry Isaac). But these will be using the two best still lenses I have, the hyper-sharp Zeiss ZF 100mm (on the 7D) and 85mm (on the GH13). Both will be at f/8 and properly focused, so any differences will absolutely not be attributable to the lenses.

Neutral on the 7D, with sharpness set to the middle. Smooth on the GH13, with sharpness set in the middle.

Whatever happens, happens. I'm just gonna go get the shots and post 'em, and y'all can decide what it "means."

P.S. I just saw your trumpet tests with GH1. It seems to alias as well! rogar-Angry(DBG): I really thought Panny did proper De-Bayering and re-scale but it's some pixel-skipping trickery...false detail and all. I just wonder why images contain so much more detail?? Are trumpets shot with new hacked firmware or OEM one? (probably doesn't matter).

 

[Kholi]

I just reviewed some footage (cable lines etc.) I took today and came to conclusion that Panny uses too much sharpening even when sharpness is at "0"...probably just to please consumer crowd. I believe the correct image on Panny should use sharpness at "-2" and noise reduction at "-2" (unless we are talking low-light). But it'll certainly be very interesting to see your test footage!

Exactly. For me, everything needs to be dialed all the way down. I said this so much before: powerlines will show it the most.

 

[Rick Burnett]

P.S. I just saw your trumpet tests with GH1. It seems to alias as well! rogar-Angry(DBG): I really thought Panny did proper De-Bayering and re-scale but it's some pixel-skipping trickery...false detail and all. I just wonder why images contain so much more detail?? Are trumpets shot with new hacked firmware or OEM one? (probably doesn't matter).

Well one problem for sure is pixel binning isn't perfect either and depending on how they decide to do it and how they interpolate data all play a huge part. if a very poor debayer is used, for computation and speed purposes, the compromises propagate. Given a bayer pattern has no two similar pixels next to each other in any fashion, they are not the best sensors for pixel binning. I have seen other sensors where they use other color filter patterns to that pixel binning can be done at the silicon level. Pretty cool stuff!

In all honesty, I wish I had more time to hack the 7D. I am working on so many projects, I just don't have the time right now. (I am an electrical engineer). I am so curious as to what the heck Canon is doing in there! With the dual processors inside the 7D, I feel like they should be able to do so much more than just allow an HD stream while recording.

 

[Grunf]

Well one problem for sure is pixel binning isn't perfect either and depending on how they decide to do it and how they interpolate data all play a huge part. if a very poor debayer is used, for computation and speed purposes, the compromises propagate. Given a bayer pattern has no two similar pixels next to each other in any fashion, they are not the best sensors for pixel binning. I have seen other sensors where they use other color filter patterns to that pixel binning can be done at the silicon level. Pretty cool stuff!

In all honesty, I wish I had more time to hack the 7D. I am working on so many projects, I just don't have the time right now. (I am an electrical engineer). I am so curious as to what the heck Canon is doing in there! With the dual processors inside the 7D, I feel like they should be able to do so much more than just allow an HD stream while recording.

Looking at trumpets here: http://www.dvxuser.com/V6/showthread.php?t=187165

Canon has quite decent horizontal behaviour. But vertical sampling is a mess. I would venture to say that Panny maybe did 2-line skip and then averaged them (they have less pixels to worry about, after all), where Canon just run with 4-line skip horizontally which gave us dancing colours due to bayer pixel-shift. Then it messes up the encoder which throws lot's of bits interpreting colour noise.

Twin CPU's is nice but I guess they are serialized and one of them just does de-bayering and other compression or something like that.
I wonder if there is horsepower left to do it "properly". Just reading raw bayer data and compressing it with wavelet might be doable, who knows.

 

[aituv]

Ok, now im 100% for canon. Just came back from a screening session on a huge screen, GH1 on big screen SUCKS. Period. I saw it with my own eyes. Looks great, detailed and everything on my 24" at home...blast it off on big screen and fail. Canon with those sensor sizes = BEAUTY.

 

[Kholi]

Ok, now im 100% for canon. Just came back from a screening session on a huge screen, GH1 on big screen SUCKS. Period. I saw it with my own eyes. Looks great, detailed and everything on my 24" at home...blast it off on big screen and fail. Canon with those sensor sizes = BEAUTY.

That's pretty convenient. Now that you've made your decision and you're one-hundred percent on that, let the rest of us actually interested in tests consider what Barry and GrunF have to show us.

Thanks.

 

[Stephen Mick]

Ok, now im 100% for canon. Just came back from a screening session on a huge screen, GH1 on big screen SUCKS. Period. I saw it with my own eyes. Looks great, detailed and everything on my 24" at home...blast it off on big screen and fail. Canon with those sensor sizes = BEAUTY.

My experience has been very different from yours.

I watched Jack Daniel Stanley and Barry Green's "Depth of Phil" on the big screen at SXSW this year, and it looked every bit as good as everything else of digital origin they screened, which included 5DII and 7D footage.

 

[adammada]

Will go out and get some side-by-side shots on GH13 and 7D. Ducks and trees (sorry Isaac). But these will be using the two best still lenses I have, the hyper-sharp Zeiss ZF 100mm (on the 7D) and 85mm (on the GH13). Both will be at f/8 and properly focused, so any differences will absolutely not be attributable to the lenses.

Neutral on the 7D, with sharpness set to the middle. Smooth on the GH13, with sharpness set in the middle.

Whatever happens, happens. I'm just gonna go get the shots and post 'em, and y'all can decide what it "means."

Thanks Barry - on it we wait everything:beer:

 

[JoeJITSU]

Ok, now im 100% for canon. Just came back from a screening session on a huge screen, GH1 on big screen SUCKS. Period. I saw it with my own eyes. Looks great, detailed and everything on my 24" at home...blast it off on big screen and fail. Canon with those sensor sizes = BEAUTY.

Unfortunately for you. You must of got hit in the face before you seen the GH1 footage.lol I have seen both camera footages and seem very indentical BEFORE the GH1 was hacked. Now with the hacked footage I believe the GH1 would improve allot more on the big screen outperforming the T2i.

 

[HDkilledFILM.]

Ok, now im 100% for canon. Just came back from a screening session on a huge screen, GH1 on big screen SUCKS. Period. I saw it with my own eyes. Looks great, detailed and everything on my 24" at home...blast it off on big screen and fail. Canon with those sensor sizes = BEAUTY.

You cannot post something like this as you have not given any reason as to why it looked so much better. Also you didn't tell us how it was screened, what compression it was output to to get on that screen etc... If you just look at this thread all the way through (which I have several times) you will see people constantly arguing that this is a TECHNICAL thread and not a personal BIAS. Different things will look better or worse to different people, the question here is which camera has the cleaner/sharper/more resolute image. If you saw the GH13 on a big screen please by all mean tell us if you have also seen the 5d mkii/7d/t2i on the big screen and why were you so impressed with it. I'm not doubting your post just trying to save you from starting a thread war.:engel017:

 

[Barry_Green]

P.S. I just saw your trumpet tests with GH1. It seems to alias as well! rogar-Angry(DBG):

Of course it does. ALL the DSLR/still cameras exhibit aliasing.

The differences are that
a) the GH1 looks a little sharper; it cleanly resolves a little more before it starts aliasing.
b) the 7D (and all the Canons) exhibits some major color contamination, which the GH1, Pentax and Nikon cameras don't.

So there's just about as much luminance aliasing in all of them. The differences are in the amount of actual resolved detail, and in the chroma contamination.

Are trumpets shot with new hacked firmware or OEM one? (probably doesn't matter).

OEM. Long, long before the hack existed.

 

[Barry_Green]

I watched Jack Daniel Stanley and Barry Green's "Depth of Phil" on the big screen at SXSW this year, and it looked every bit as good as everything else of digital origin they screened, which included 5DII and 7D footage.

And, I might point out, that was before the hack. That was shot on an early-model original GH1.

 

[Ian-T]

Well...I've always said the GH-1 seemed sharper than the 5Dll...and that was a year ago when it first came out. Now with this hack...there's no question. But as most of us already know sharper doesn't mean better. What makes one better is the operator behind the cam (whichever one they are using at the time). I got to admit...technically speaking of course...the Panasonic is now better than the Canons. Aesthetically...5Dll is still the shiznit....that isuntil someone does a Phil Bloom on the Panasonic...which I'm sure is right around the corner. If only Panasonic did it right the first time then many of us would be settled today. Anyways...It's great to have so many different kinds of choices.

Now....I gotta go shoot someone.... with a camera that is.