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

One of the reasons that birders like APSC is the field of view in the viewfinder. In a 5D/6D body, with a 400-600mm lens attached, a small bird is still very small. It might turn out ok in crop, but it's hard to tell if you're even in focus through the viewfinder with a full frame camera, without going to magnification liveview, even when you have enough pixels to make a decent photograph.

For me, one of the real benefits of a high res mirrorless is that if you can toggle between crop and full frame in the EVF with the push of a button, so you can use the wider field of view to compose -- or just find your subject -- and the cropped field of view in the EVF (and further magnification if you want it) to focus, or to simply end up with much smaller files instead of cropping in post. Plus, you can't shrink the original RAW by cropping.

Obviously, we have no idea how the high resolution Canon sensor will perform. But here's the issue with high-resolution sensors, to date -- they just aren't a replacement for having longer (or ideal) focal length lens, but it's sure tempting to try to do so. A MFT 300mm can be razor sharp, and yet the photos are not as pleasing as a 1DX with a 600mm at a much lower resolution. Or to put it another way, if you deeply crop 3 megapixels out of a 50 megapixel photo, it won't look remotely as good as if you had enough focal length to take 50 megapixels, and then reduced that to 3 megapixels.

Olympus MFT is the system I had after I switched from Nikon and before I came to Canon. One of the problems for it is that there simply are no 400mm-600mm+ lenses, and you can't get to those focal lengths without shrinking the aperture to way too small. In real terms, I don't have an issue with a full sized body and a 100-400 II lens mounted -- I can hike and shoot with that for 10+ hour stretches, and the results I get are just superior than an MFT with a 300mm, especially in bird in flight photos.

The reason I left Oly wasn't that I didn't like their cameras -- it's that I felt like I hit a ceiling with BIFs for even large-sized birds (like heron), where the keeper rate was significantly lower than Nikon, and I was frustratingly unable to increase it. Of course, that was a long time ago, and perhaps things have changed -- but now, I am a very happy Canon shooter, and for my non-birding stuff, Canon just has a much more complete lens portfolio with good options in a spectrum of price points. Plus, the third party accessory ecosystem is second to none.

Putting price aside, at 400mm and above, and especially with mirrorless, there is no advantage to MFT or APSC for the body, since the lens does not get any smaller with a smaller sensor, and I personally find that a larger body acts as a better balance to the heavy glass.

Anyways, all this to say, I'm excited to see how Canon's 100 megapixel RF performs. If it offers the kind of leap that the 5DSR did, I will probably buy one.
The reason for the disappointment in µ43 images is because of the small pixel/high density issue. Any slight shortcoming will be magnified more than the FF equivalent. Just as MF prints looked so much better than 35mm prints. At small sizes and excellent technique (aided in no small part by Oly's excellent IBIS) µ43 looks splendid. My 5DsR looks splendid at FF and when cropped in close still looks great. But only with tripod mounting, careful focus and strobe.
So going to 100MP will yield an all-in -one package but that added crop mode may be far less useful than one might hope.
 
Upvote 0

dtaylor

Canon 5Ds
Jul 26, 2011
1,805
1,433
You're probably mixing up the MTF/resolution and 'information'. Again sharpening may increase visual sharpness, but it never actually recovers anything.

No, I'm not confused about information, and yes, sharpening and other image processing techniques can recover "lost" information. This is known and discussed in scientific fields where the information is actually of importance.

For us, for all practical intents and purposes, the resolution difference between AA/no AA cameras like the 5Ds/sr and D800/800E is not even observable, much less observable in print. What we observe is the MTF curve shift, the slight loss in contrast, something easily changed by light sharpening with no negative impact.
 
  • Like
Reactions: 1 user
Upvote 0

dtaylor

Canon 5Ds
Jul 26, 2011
1,805
1,433
The 300 Oly does have the same AOV as the FF 600. No need to make the 600.

Unless, of course, you want 1200mm equivalent.

Your proposal is exactly the model of the µ43 but with the added bulk and expense of FF optics.
The only "advantage" being that at FF one would now have 100MP, a dubious advantage.

The two advantages would be 100mp when using the FF, which is of significant advantage for some subject/view size combinations, and far better high ISO performance when using the FF. One camera covering multiple uses and angles. Still, I could see some people wanting an APS-C or m43 system, with the same pixel density, just to save weight on shoots where reach is the only concern.

The 5Ds replaced APS-C crop for me in long telephoto scenarios. If I have to crop it's still just as good. And in that moment when I don't have to crop...when I'm able to get close enough to the subject...it's off the charts good.
 
Upvote 0
Dec 31, 2018
586
367
Now that camera what is on harrys desk must be R5 with 100mpx, because i got one lot more powerfull on my desk, i guess it must be R1.
One i got is 150mpixel 16k video able and it got another sensor too 30mpixel one .
Both move sideway rails under photo plane and claws of ibis locks to wanted one and raises it to right place.
Shooting speed is 5fps/sec with 150mpx and 15 fps/sec with 30mpx.Nothing spectacular but there isnt all!!
Inside neck strap goes data transfer cable from camera and behind neck on strap is socket.
Quess what you plug there :) Cable from TURBO unit!!
Its 20cmx10cmx5cm box what contains slot for big battery ,10 slot for memory cards and 10x digic 9 processor.
You can place it to backpack when its connected to camera. weighting like 1kg
Basically it can edit and write simultaneusly 10 different picture cause 10 processor and 10 card.
So it makes camera shoot 10x faster,50fp/s with 150mpx and 150 fp/s with 30mpixel,And buffer capasity is bigger too.

More about camera . Canon listened whines about how it is wasting real estate on back and upside,so now there is only onn/off button,touch bar and nothing else!!
What amazing save of real estate!!
Touch bar is now more and less adjust button.if you click shortly left side it decreases 1/3 stop ,if clicking more long left side it decreses 1 stop
If you slide from middle to left side it decreases 5 stop.You can do it fast with right thumb when left thumb adjust ring on objective and all other clumsy fingers just support camera.
In viewfinder there is little cameras what track eye movements like topgunners helmet.Bottom of backside there are also cameras what track about mouth movements.
So when you want move focus point ,opening mouth activates focus point change and all you need do is look point you want lock focus and then close mouth.
Or you can left mouth open and camera focus everywhere you look but it may not be good idea cause eyes wanders to look framing too.To shoot all you need do is stick tongue
out from mouth to middle and it shoots on burst untill you pull tongue back.
When you want change shutter speed you activate changing mode when sticking tongue out to right cheek and you change value with touch bar.
When sticking tongue straight to left cheek you activate iso changing and can adjust it from touch bar too.
All values can be changed middle of shooting too when moving tongue from middle to edge of mouth.
When you show your teeths you activate menus and you can navigate there with eye locking and tongue stick clicking ,so viewfinder does all jobs what oldfashioned touchscreens did.
Situtations where you cant look to viewfinder there is virtual reality glasses what allow you see viewfinder where ever camera is ,all eye and mouth commands works with glasses too.
 
Upvote 0

dtaylor

Canon 5Ds
Jul 26, 2011
1,805
1,433
Sorry didn't see your message yesterday as this forum doesn't show notifications on mobile view. Anyway, my point is, as far as I'm aware, all known sharpening methods are irreversible, you can't apply the inverse algorithm and get the original blurred image back exactly as it was.

I'm almost certain this is false, depending of course on algorithm and parameter range. But sharpening by convolution should be reversible within a certain parameter range.

I agree AA filter does little harm but still does, it's evident when you compare samples from 5DS with and without the filter.

With light sharpening it's nearly impossible to discern 5Ds/sr images or D800/800E images, even while pixel peeping. I would agree that a computer could still tell them apart, but that's because the AA filter has a varying impact depending on detail contrast. So if you sharpen to match high contrast details then low contrast details on the sharpened AA filtered image are sharper than on the non-AA image. (Yes, I investigated it to that depth.)
 
Upvote 0

dtaylor

Canon 5Ds
Jul 26, 2011
1,805
1,433
It’s also going to be an average (or poor) performer in low light having to cram that many pixel sensors on the chip. The more pixels the smaller the color sensors have to be to fit on the chip, which is why low megapixel cameras like the a7s can practically see in the dark.

For stills photography higher resolution sensors are observably better at high ISO, when compared at the same view size. (When pixel peeping the flaws in the higher resolution image are magnified more giving a false impression of performance.) At the same view size they have similar noise characteristics but better sharpness and detail.

"5Ds is not a high ISO camera" was a review meme that was demolished by Imaging Resource in their print tests, and I'm convinced Canon capped the max ISO for marketing reasons (i.e. because of pixel peepers). You can also prove it false by using dpreview's comparison tool using the same view size mode. The A7s does not turn in a very good performance compared against any higher resolution FF sensor. Canon, Nikon, or Sony...give me their highest resolution FF chips for high ISO stills.

Video is a different beast because we don't have processing engines that can optimally scale high resolution frames to 4k or 1080p at 24-60 fps. For video the optimum high ISO sensor might indeed be one that is resolution matched to the intended output.
 
  • Like
Reactions: 1 user
Upvote 0
Dec 13, 2010
4,932
1,608
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.
“It certainly isn’t because they can’t “ can you back that claim up with something?
 
  • Like
Reactions: 1 user
Upvote 0

Talys

Canon R5
CR Pro
Feb 16, 2017
2,127
451
Vancouver, BC
So going to 100MP will yield an all-in -one package but that added crop mode may be far less useful than one might hope.
I agree. This is why I say, I'm not sure if what I want is a pipe dream or a reality that's just a year or two away :) I suspect that the 100mp camera will be somewhere in between, where it's not like I'll be able to throw on a 200 or 300mm lens and call it a day -- but where I'll be able to get better, deeper crops and where some photos will come out better than they would with 30 - 50 megapixel cameras. If that's the case, I'll probably still buy it.

On the other hand, I will say that when using the Sony a7r3, I thought that the crop mode was one of its nicest features, and that the APSC crops of 40-something megapixels was, situationally, a genuinely useful feature.
 
  • Like
Reactions: 1 user
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
I think a 100 MP 35mm sensor camera will need a general freeze-down of the world to make full use of those tiny pixels...
Those pixels are really not all that tiny.

I don't own one, but from what I can see online the Canon G7x II is not considered to be hot garbage... And it has far smaller pixels! If you would scale that 1 Inch 20MP sensor (116mm^2) up to full frame size (864mm^2) that would result in a resolution of 149 MP.

If people can get good shots today with that old sensor of that density, I can't understand why some people here assume somebody would have to do something they don't do now, like freezing the world or shooting only on a tripod, when using a sensor with the SAME density and newer, likely superior sensor technology in the future?

I'm not mad about it :), it just seems weird to bring up such points when other concerns, like big file size or limited framerates are far more reasonable in my eyes.
 
  • Like
Reactions: 1 users
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
I'm almost certain this is false, depending of course on algorithm and parameter range. But sharpening by convolution should be reversible within a certain parameter range.
Wikipedia hasbthe following to say about that:

"The usual method is to assume that the optical path through the instrument is optically perfect, convolved with a point spread function (PSF) [...] If this function can be determined, it is then a matter of computing its inverse or complementary function, and convolving the acquired image with that. The result is the original, undistorted image.
In practice, finding the true PSF is impossible, and usually an approximation of it is used, theoretically calculated[4] or based on some experimental estimation by using known probes. [...] The accuracy of the approximation of the PSF will dictate the final result. Different algorithms can be employed to give better results, at the price of being more computationally intensive. Since the original convolution discards data, some algorithms use additional data acquired at nearby focal points to make up some of the lost information."

So it looks to me that it is wrong to say sharpening using deconvolution can't bring back true detail. It's also right to say that what is brought back is not the 100% identical to the non convolved original in real world application.

But at that point, it seems to become splitting hairs for most cases to me. After all, the true original subject is approximated anyway through digital imaging by beeing measured in discrete pixel steps. And without an AA filter, there are also occasionally occuring artefacts - Moiree is also a mistake in the imaging process that causes the resulting image to differ from the true original. Conceptionally, where is the difference between that and sharpening artefacts?

So keeping practical purposes a priority, I'm inclined to believe the people who have done the testing and say that the AA filter can be effectively canceld in post through mild deconvolution sharpening. That's just the impression I've got from reading though, so it doesn't matter anyway.
 
  • Like
Reactions: 1 user
Upvote 0
"I don’t think we’ll see such a camera until 2020, as I believe there need to be some more native RF lenses announced to handle landscape and studio shooters."

I'd expect a 4Q 2019 release so it can attempt to compete with the GFX-100s (in resolution at least). But I don't think the hold up for such a product will be a drought of native lenses; Canon would certainly prefer to have more lenses for the RF on the market, and it seems they plan to do exactly that in 2019, but there's also plenty of L-series EF lenses that can take full advantage of a 100MP sensor at wide and moderate apertures.

I think the hold up for this camera may just come down to delivering a highly capable IBIS system. A lot of the top-notch RF glass at this point doesn't have OIS, and you would really need some form of image stabilization to generate sharp 100MP images at flash sync shutter speeds (presumably 1/180th second). I'd be happy to wait several more months for its release if Canon can deliver a hybrid OIS/IBIS system like Oly's that uses both lens and sensor stabilization for 6+ stops of correction.
 
  • Like
Reactions: 1 user
Upvote 0

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,355
22,528
Wikipedia hasbthe following to say about that:

"The usual method is to assume that the optical path through the instrument is optically perfect, convolved with a point spread function (PSF) [...] If this function can be determined, it is then a matter of computing its inverse or complementary function, and convolving the acquired image with that. The result is the original, undistorted image.
In practice, finding the true PSF is impossible, and usually an approximation of it is used, theoretically calculated[4] or based on some experimental estimation by using known probes. [...] The accuracy of the approximation of the PSF will dictate the final result. Different algorithms can be employed to give better results, at the price of being more computationally intensive. Since the original convolution discards data, some algorithms use additional data acquired at nearby focal points to make up some of the lost information."

So it looks to me that it is wrong to say sharpening using deconvolution can't bring back true detail. It's also right to say that what is brought back is not the 100% identical to the non convolved original in real world application.

But at that point, it seems to become splitting hairs for most cases to me. After all, the true original subject is approximated anyway through digital imaging by beeing measured in discrete pixel steps. And without an AA filter, there are also occasionally occuring artefacts - Moiree is also a mistake in the imaging process that causes the resulting image to differ from the true original. Conceptionally, where is the difference between that and sharpening artefacts?

So keeping practical purposes a priority, I'm inclined to believe the people who have done the testing and say that the AA filter can be effectively canceld in post through mild deconvolution sharpening. That's just the impression I've got from reading though, so it doesn't matter anyway.
You might find this article interesting, much better written than in Wiki: https://www.microscopyu.com/techniq...the-diffraction-barrier-in-optical-microscopy
It does contain the statement: "Spatial frequency information that is lost during imaging cannot be recovered"
 
Upvote 0

SecureGSM

2 x 5D IV
Feb 26, 2017
2,360
1,231
It will, but perhaps not the way most people are thinking...

Take a 100Mpixel image, downsize it to 25, and gain two stops of DR.

hard clipped highlights cannot be recovered past certain point downsampled or not. you gain sharpness and extra shadow recovery through noise suppression.
 
  • Like
Reactions: 1 user
Upvote 0
Unless, of course, you want 1200mm equivalent.



The two advantages would be 100mp when using the FF, which is of significant advantage for some subject/view size combinations, and far better high ISO performance when using the FF. One camera covering multiple uses and angles. Still, I could see some people wanting an APS-C or m43 system, with the same pixel density, just to save weight on shoots where reach is the only concern.

The 5Ds replaced APS-C crop for me in long telephoto scenarios. If I have to crop it's still just as good. And in that moment when I don't have to crop...when I'm able to get close enough to the subject...it's off the charts good.

Yeah, like everyone, 1200mm is the one I am always reaching for.

When I got my DsR for architecture clients I thought that it would be a quantum step up i IQ. What it was, was a step up in processing time with excellent but not useful improvements in resolution. Side by side comparisons of tripod mounted shots at 30x40" showed a slight but nearly indiscernible improvement in sharpness. Going larger shows the difference more clearly but the proper viewing distance eliminates the difference.
Then some clients whined about getting massive files.
 
Upvote 0
“It certainly isn’t because they can’t “ can you back that claim up with something?
Google is your friend here. Canon's highest density sensor has a pixel size of 4.1 microns. Sony has an entire line of sensors with much smaller pixels all the way down 1.0 microns. How many 1.0 micron pixels you could put on a full frame sensor? I'm not going to do the math but it's a lot more than 100 million. I like Canon's system and I'm no Sony fan but you are kidding yourself if you think Canon is ahead of Sony in sensor development.

edit: sorry, forgot about the sensor in the 80D. That one is 3.7 microns.
Also important to remember that it's a linear measurement so Sixteen 1-micron pixels fit into the same area as One 4-micron pixel. Fair to say that Sony is pretty good at small pixels.
 
Last edited:
Upvote 0

Don Haines

Beware of cats with laser eyes!
Jun 4, 2012
8,246
1,939
Canada
Google is your friend here. Canon's highest density sensor has a pixel size of 4.1 microns. Sony has an entire line of sensors with much smaller pixels all the way down 1.0 microns. How many 1.0 micron pixels you could put on a full frame sensor? I'm not going to do the math but it's a lot more than 100 million. I like Canon's system and I'm no Sony fan but you are kidding yourself if you think Canon is ahead of Sony in sensor development.

edit: sorry, forgot about the sensor in the 80D. That one is 3.7 microns.
Also important to remember that it's a linear measurement so Sixteen 1-micron pixels fit into the same area as One 4-micron pixel. Fair to say that Sony is pretty good at small pixels.
If you look at the 1/2.3 sensors, scaling up a 16Mpixel sensor to FF size gets you 485Mpixels. I think that everyone and their dog has the ability to pack lots of pixels on a FF sensor.
 
  • Like
Reactions: 2 users
Upvote 0