5D MKIII specs


Downloaded the 1080p file.

-moire looks fixed
-aliasing looks fixed

Can't tell anything about rolling shutter, but my prediction would be that it's improved quite a bit, but still there if you look for it.

However

-resolution looks the same as 5D II (considering the processor and the sensor resolution being a multiple of 1080p... how do you screw this up Canon???)
-Noise looks like traditional ugly Canon noise on the last generation of DSLRs. Damn. Why can't it look like grain similar to how the GH2 renders it????

Somewhat disappointing. Probably not upgrading. 5DII + ML + VAF5DII seems equally good.

Wide angle shots looks a tad sharper than 5D II, but it's not like you can't already achieve this with ML. Ok, maybe a tad better, but that's just hair splitting honestly.

As a 5D II user with a VAF5DII, I'm contemplating just buying a GH2 and a Samyang 24mm f1.4 that I can use to swap between cameras, and then a Samyang 14mm f2.8 for the same reason. I guess down the road I could get that overpriced electric Canon-Panasonic adapter thing that has Autofocus so I can take advantage of my Canon lenses. That seems to be pretty cool.

Man, if the 5D III had GH2-like resolution (with less digital sharpening) and the moire/aliasing/rolling-shutter fixed, I would be all over it.
 
Last edited:
question on the multiple theory. doesnt the c300 shoot 4k then resamples to 1080?
why are we expecting more if the mk3 doesnt do that?
 
the 5D3 shoots 5760x3240 and then resamples to 1920x1080
it's a clean 3x downsampling
I can't believe they managed to get it to look so soft
it's probably still reading only a third of the lines, then applying crazy math to avoid aliasing/moire, otherwise it just doesn't make sense that it looks SO soft
 
question on the multiple theory. doesnt the c300 shoot 4k then resamples to 1080?
why are we expecting more if the mk3 doesnt do that?
As I understand it, the C300 is a 4K sensor, but samples the RGB values as two green pixels, one blue, and one red. So four pixels are sort of "binned," but with real (not interpolated) 4:2:2 color, at 1,920 x 1,080.
 
I noticed you referred to it as a 4k sensor, to get technical it's not quite 4k ... its Quad HD which is 3840 wide 2160 high.

This means its exactly four x 1080p. Two green 1080p's ... one red 1080p and one blue 1080p.

Two green pixels + 1 red + 1 blue debayers to 1 RGB pixel.

Source: This information was projected during the c300 presentation on Nov 3rd
 
They'r not binned ... those 4 represent 1 RGB pixel.
I used the term binned just to help explain that a group of four is sampled together as one RGB pixel, but you're right. Three different channels are read, so it's not really one big pixel made up of 4 smaller ones, so it doesn't increase sensitivity while reducing noise (at the cost of resolution) the way true binning does. And, yes, I should've said the sensor is actually QuadHD (3,840 x 1,080), not what's conventionally known as 4K (4,096 pixels wide).
 
I used the term binned just to help explain that a group of four is sampled together as one RGB pixel, but you're right. Three different channels are read, so it's not really one big pixel made up of 4 smaller ones, so it doesn't increase sensitivity while reducing noise (at the cost of resolution) the way true binning does. And, yes, I should've said the sensor is actually QuadHD (3,840 x 1,080), not what's conventionally known as 4K (4,096 pixels wide).

I think you mean 3840x2160 for QuadHD.

5760 x 3840 is the max resolution (let's assume movie mode just used 3240 for a 16:9 aspect ratio), here is the problem I see:

RGR GRG RGR GRG
GBG BGB GBG BGB
RGR GRG RGR GRG

If they used 9 bayer pixels to recreate each final image in the most simplistic of algoritms, look how unbalanced the sampling is:

pixel 1 = 4G - 4R - 1B
pixel 2 = 5G - 2R - 2B
pixel 3 = 4G - 4R - 1B
pixel 4 = 5G - 2R - 2B

The advantage of R or B depends on whatever the first row is (I just made a guess it was RGRGR)

3x pixels to me is not an easy to downsample bayer pattern multiple of the resolution because we don't have and equal number of RGB so it's really RGGB which is why the 4x of the C300 makes the most sense (and easiest down convert)

I don't know what method Canon is using to downsample, but to get rid of the moire/aliasing of the 5dm2 it has got to be a lot more complex than people are thinking.

(1) The key is the downsample. It cannot be the same as the 5Dm2 because if it were, the same image magic post downsample we could do in post.

(2) It's possible they are sacrificing false detail to get rid of the moire/aliasing. This means, the image IS actually sharper than the 5Dm2, but you are actually seeing the resolution without false detail, which you couldn't do in the 5dm2. Of course, if you think it's too soft for you, that's valid as well.

The point I am making is that the image might be vastly improved but not look as much because of improvements to false data.
 
What's everybody complaining about? A 5dII with no moire, aliasing and an in camera 90Mbit intra codec.

If that's not good enough for you then your obviously producing for the cinema in which case you should buy a C300 assuming you want a canon that is :evil:.

It was never going to have the resolution of the C300, nowhere near. The sensor is designed for great STILLS as it's a STILL camera. The c300 has a sensor designed for VIDEO and it takes great video at that.


And before anybody brings it up, the GH2 does not take GREAT stills..
 
What's everybody complaining about? A 5dII with no moire, aliasing and an in camera 90Mbit intra codec.

If that's not good enough for you then your obviously producing for the cinema in which case you should buy a C300 assuming you want a canon that is :evil:.

It was never going to have the resolution of the C300, nowhere near. The sensor is designed for great STILLS as it's a STILL camera. The c300 has a sensor designed for VIDEO and it takes great video at that.


And before anybody brings it up, the GH2 does not take GREAT stills..

Way back when, in still photography, we could buy the same film that pros used.
 
I actually will use the 5DmkIII to shoot stills with when I get it in three weeks and I'll keep the mkII as a second body. I have a AF101 which is a much better tool when it comes to video. I see the improved video features as a bonus which might come in handy but not essential.
 
I actually will use the 5DmkIII to shoot stills with when I get it in three weeks and I'll keep the mkII as a second body. I have a AF101 which is a much better tool when it comes to video. I see the improved video features as a bonus which might come in handy but not essential.
+1, as i will be in that exact situation.
 
Way back when, in still photography, we could buy the same film that pros used.

And you can still (now that that we don't use film) buy the same camera the pros use. The 5D mk3 (borderline), the 1DX mk3, D3X etc... are certainly for sale. So are the red Epic, Arri Alexa, etc... on the video side. We're talking about prosumer video use of a hybrid DSLR.
 
Back
Top