IBIS and 100mp coming to an EOS R camera? [CR2]

Jul 12, 2014
293
270
The AA filter hardly makes a difference between 5Ds and 5Dsr images while pixel peeping. Except of course when moire rears its ugly head, at which point the difference is observable at normal sizes unless/until you clean up the 5Dsr shot.

Canon dominates wedding and fashion and has resisted the "no AA filter" meme. I hope they continue to do so, or at least offer the choice on their highest resolution bodies.
I beg to differ. An AA filter diminishes resolution of fine detail by 10% +/-. This is why the higher priced 5DSR outsells the 5DS and why other camera manufacturers have all but abandoned the AA filter. To realize the most benefit of a 50mp sensor requires the finest/expensive lenses, so why give back resolution in every single image just to avoid very rare moire, which is easily corrected in post should it appear. Canon has done well among wedding photographers, not because of the AA filter, but because of the advances offered by the 5D2 in its day and the family of Canon lenses. Having made the investment, switching ecosystems is expensive.
 
Upvote 0

Sporgon

5% of gear used 95% of the time
CR Pro
Nov 11, 2012
4,719
1,537
Yorkshire, England
I beg to differ. An AA filter diminishes resolution of fine detail by 10% +/-. This is why the higher priced 5DSR outsells the 5DS and why other camera manufacturers have all but abandoned the AA filter. To realize the most benefit of a 50mp sensor requires the finest/expensive lenses, so why give back resolution in every single image just to avoid very rare moire, which is easily corrected in post should it appear. Canon has done well among wedding photographers, not because of the AA filter, but because of the advances offered by the 5D2 in its day and the family of Canon lenses. Having made the investment, switching ecosystems is expensive.

No it doesn’t. Before I settled for the 5DS I tried both s and sr together. Obviously from unsharpened raw the sr is sharper with greater contrast, and, if we’re going to be anal, occasionally some colour aliasing. Equally obvious is the fact that from an unsharpened file the sr will produce a higher “resolution” MTF when compared with the unsharpened s. But apply a tiny amount of USM and the difference has gone.

Long live the AA filter on a Bayer Array sensor.
 
  • Like
Reactions: 2 users
Upvote 0

sdz

CR Pro
Sep 13, 2016
262
209
Pittsburgh, PA
I think we are very close to minimal additional improvements.The conversations about noise and DR are somewhat overblown as the differences seem to have small impact. It is hard to believe that everyone needs to stretch their files into such unrecognizable shape that magical DR (that actually doesn't exist) will save their lives.

Marginal utility explained

Greater than marginal improvements in dynamic range would likely require a new sensor technology and great improvements in computing power, power management, etc. I'll not hold my breath waiting for these technical leaps.
 
Upvote 0

sdz

CR Pro
Sep 13, 2016
262
209
Pittsburgh, PA
Canon seems to be the only team left in the Megapixel wars. If 100MP sensors significantly outperform 45MP sensors why doesn't Sony use them in it's cameras or manufacture and sell them to it's sensor customers. It certainly isn't because they can't. Dividing up the Pizza into smaller and smaller pieces doesn't create more pizza.

Sony supplies sensors to Phase One.
 
Upvote 0
Even for landscapes, 100 mp seems to be an overkill. 50mp would be just fine.
However if say a 100 mp camera will be able to downsample to 25-33mp with DR increase, it'd be very nice.
Not overkill if you regularly want to make large prints ;-) Downsampled 5Ds images look good, so yes 100+MP would interest me...
 
Upvote 0
It would be interesting to see how something like this performs in real life. Panorama files from a 5DSr are already nearing un-manageable status.

If I shot a 17 frame panorama with a 100mp camera, I would have no way to process it.

I would rather see Canon release a 25-30 mp sensor based on Foveon technology.

Unmanageable? Hardly, I regularly include dozens of 5Ds images in high res survey panos, and I'm working on a 2010 Mac Pro
 
Upvote 0
Jun 9, 2017
124
18
When people think about 100mpx being too much, they forget about the bayer filter that essentially means the accurate colour resolution of an image is 1/4 of the total resolution, because each of the 100 million pixels on the sensor is only receiving either red, green or blue light so the colour information from four pixels (at least, more in more complex algorithms) is required to generate the colour for a single pixel. Generally this isn't a big problem, but if you have images with large areas of predominantly one shade, such as a green plant or a red dress, the real perceptive resolution of what that sensor will provide will be significantly less than the 100 megapixels.

So, downsampling a 100 megapixel image to a 25 megapixel image would give you a 25 megapixel image with, assuming an ideal lens, the optimum sharpness and colour accuracy.

Is this important for most people? Of course not. But there are certain tasks, such as high resolution reproduction of artworks or natural objects, where capturing the finest detail accurately is important, and for that a 100 megapixel EOS R can't come soon enough!

Of course at 100megapixels diffraction is going to be a real bitch, but that's another thing altogether!

Ehm, am I completely mistaken or are you? I thought a pixel consists of 4 (or 3) subpixels that have the colorfilters in front of them. So if this 100 mpx is not marketing speech bullshitting the buyer than it should mean that we have indeed a sensor at hand with 100 mpx that can display color (meaning 400 m-subpx).

When I open raw files from my 26 mpx 6D mk II, the files have 26 mpx (as advertised) independent of the scene I took the image of. I definitely don‘t have a 6,25 mpx file...

Am I missing something?
 
  • Like
Reactions: 1 user
Upvote 0
No it doesn’t. Before I settled for the 5DS I tried both s and sr together. Obviously from unsharpened raw the sr is sharper with greater contrast, and, if we’re going to be anal, occasionally some colour aliasing. Equally obvious is the fact that from an unsharpened file the sr will produce a higher “resolution” MTF when compared with the unsharpened s. But apply a tiny amount of USM and the difference has gone.

Long live the AA filter on a Bayer Array sensor.
Physical AA filter and software sharpening filter only cause loss of information, and sharpening doesn't actually recover anything, it sacrifices more information to make the image sharper visually. Measuring MTF after applying a software filter that artificially increases MTF doesn't make much sense.
 
Upvote 0
No it doesn’t. Before I settled for the 5DS I tried both s and sr together. Obviously from unsharpened raw the sr is sharper with greater contrast, and, if we’re going to be anal, occasionally some colour aliasing. Equally obvious is the fact that from an unsharpened file the sr will produce a higher “resolution” MTF when compared with the unsharpened s. But apply a tiny amount of USM and the difference has gone.

Long live the AA filter on a Bayer Array sensor.
Definitely - I've compared the two and am quite happy I decided to save the couple of hundred quid difference when I got the 5Ds

'Lack of AA filter' seems largely a marketing led feature to extract more from the punters and perhaps also those relying on some maths over actually taking some photos to look at ;-)

Tried the Hasselblad X1D-50C recently and when moire hits with that you -really- notice it - otherwise really nice
 
  • Like
Reactions: 1 users
Upvote 0

Don Haines

Beware of cats with laser eyes!
Jun 4, 2012
8,246
1,939
Canada
This is true. My Olympus produces bitingly sharp images. I only have the Olympus M. Zuiko Digital ED 12-40mm f/2.8 Pro. Don, do you happen to have the 40-150mm? I think that's my next step.
We just have the 12-50 and the 40-150. They are ok, but not as nice as the pro lenses. The whole reason we got an Oly was to have something that fit into a tool bag.
 
  • Like
Reactions: 1 user
Upvote 0
funny.... I remember all the same arguments when we left 8 megapixels behind......
Indeed, some of the film/digital ones before are quite amusing to look back on too ;-)

Personally, I might appreciate the mathematical analysis, but when it comes down to it, it's taking photos and making your own mind up that wins out...
 
Upvote 0

dtaylor

Canon 5Ds
Jul 26, 2011
1,805
1,433
Physical AA filter and software sharpening filter only cause loss of information,

You're applying binary thinking to a physical phenomenon which is not binary. An AA filter shifts the MTF curve slightly. In theory that should result in a loss of resolution (ability to distinguish line pairs) at the very extinction limit of the MTF curve. In practice you would be very hard pressed to observe the loss even while pixel peeping a resolution chart because contrast at extinction is so low. The difference you can actually see with your own eyes occurs at higher contrast levels. It's also a difference which can be mitigated through software.

Note that analysis software which takes a resolution chart image and spits out a single number does so based on a specific contrast point that is close to but above actual extinction. You can sharpen the AA filter image before feeding it to the software and get the same output resolution number. By the same token you can, to a point, sharpen either image to boost the output numbers. In other words: the software's output is an imperfect approximation of reality, not the gospel truth. And it couldn't be any other way since resolution is not a single number. It's a curve.

and sharpening doesn't actually recover anything, it sacrifices more information to make the image sharper visually.

Sharpening does not sacrifice image data (except perhaps in extreme cases or with poor algorithms) and can in fact allow recovery of some information.
 
  • Like
Reactions: 1 user
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
Physical AA filter and software sharpening filter only cause loss of information, and sharpening doesn't actually recover anything, it sacrifices more information to make the image sharper visually. Measuring MTF after applying a software filter that artificially increases MTF doesn't make much sense.

Edit: Phone messed up and hit reply before I could type anything...

First up, I don't fully understand the details myself, but I've been interested in the topic for a few weeks now and did some basic research. With that said, as far as I know you're assumptiin of information getting lost because of the low-pass (AA) filter in cameras is partially wrong. Some Information may get lost, but the majority is just convolved with information from the sourrounding pixels. Through a technique called deconvolution, this information can be restored. The following PDF document and webpage have some technical explanations and nice examples of what can be achieved:
- More general and technical: https://www.google.com/url?sa=t&sou...Vaw13sfg4ciYMGPUqW6g_u3k6&cshid=1547385810817
- Easier to read and nice visual examples: http://www.clarkvision.com/articles/image-restoration2/

The point is, for small, predictable amounts of blur like an AA filter produces, the amount of truly lost information is probably neglible. It's just a matter of applying the techniques to get the Information when it's needed.
 
Last edited:
  • Like
Reactions: 1 user
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
Ehm, am I completely mistaken or are you? I thought a pixel consists of 4 (or 3) subpixels that have the colorfilters in front of them. So if this 100 mpx is not marketing speech bullshitting the buyer than it should mean that we have indeed a sensor at hand with 100 mpx that can display color (meaning 400 m-subpx).

When I open raw files from my 26 mpx 6D mk II, the files have 26 mpx (as advertised) independent of the scene I took the image of. I definitely don‘t have a 6,25 mpx file...

Am I missing something?
You can take the marketing numbers literally. A 24 Megapixel sensor has 12 Million green pixels, and 6 million red and blue ones. And talking about the resolution of screens like the LiveView and ViewFinder Canon at least also calls it 2.1 Million dots when they should say 0.7 million dots of each color (or 0.7 Megapixel, a.k.a not even HD).

By the time you open a raw file in any editor, a step called debayering will have been applied to it. For soc JPEG images this is done in camera. This will overlay information extracted from the sourround pixels over each of the single colour pixels to estimate the two missing components to form an actual pixel in the digital sense.
 
  • Like
Reactions: 3 users
Upvote 0
Sony supplies sensors to Phase One.
The Phase One sensors are 2.5 times larger than Canons full frame sensors. The 150 MP XF IQ4 has a pixel density similar to a 60MP full frame. The 100 MP XF 1Q3 would be equivalent to 40MP. Fuji's GFX 100S has a smaller sensor so the density will be similar to the IQ4. (ie. bigger sensors can support more pixels).

The physical properties of light dictate that there are limits to the benefits of smaller pixels and that we will soon reach a point where further miniaturization is counter productive. Short of going to larger sensors, a good strategy would be optimize existing sensors through multi-sampling, computation and pixel shifting which is where Sony seems to be heading.
 
Upvote 0
Mar 2, 2012
3,187
542
Ehm, am I completely mistaken or are you? I thought a pixel consists of 4 (or 3) subpixels that have the colorfilters in front of them.

Nope, the previous commenter is correct. Some
Canon sensors have subpixels (those with DPAF), but that is neither here nor there. In traditional sensors, each pixel is only sensitive to a narrow passband: red, green, or blue.

When I open raw files from my 26 mpx 6D mk II, the files have 26 mpx (as advertised) independent of the scene I took the image of. I definitely don‘t have a 6,25 mpx file...

Am I missing something?

What you’re missing I think is that you’re looking at a debayered depiction of the raw. It has 26 million pixels, and it also has more than three colors, yes?
 
Upvote 0
May 4, 2011
1,175
251
We’ve been told that Canon scrapped the development of an EOS 5DS/R follow-up and that their next high megapixel camera will be for the EOS R system.

If true, too bad. Although that said, the current model is good enough most likely to get several more years out of...

As for the 5DS vs. 5DSR discussion, I suspect that the filter in the regular 5DS is relatively weak, as I remember when I tried it out I was still impressed by the clarity and sharpness of the images in test shots. I went for the R version because I wanted every last bit of detail (especially at that resolution anyway), but the S was still putting out crisp results if I recall. However, the difference is much more apparent when compared to either the 5D3 or 5D4. 5DSR shots often look crisper, sharper and have more pop to them (even with the same lens!)...of course much of that can be attributed to the additional MP, but I'm sure the lack of AA filter helps too. Especially when the 50MP looks crisper than the 30MP one when both are viewed at 100%...
 
Upvote 0