Rumored Olympus OM-D E-M5 successor may offer resolution-boosting 'sensor shift'

I am not bashing Olympus; just making a half-joke with probably some truth to it. I like (and use) various brands, including Olympus. Each has their strengths and weaknesses.

Olympus is on record as saying that 16MP is the sweet spot for the enthusiast sector ... if they offer sensor shift tech for more mp then it is a bonus - not something to ridicule ... even now the E-M5 is still more camera than most can handle. When your gear list is more awe-inspiring than your photographs then its time to step back and reassess... I've seen amazing photos in some of the challenges using even high-end P&S cameras.

GWYNOXY, amazing photos can be made on phones too, but to convince a higher-end enthusiast to spend hundreds on a camera is a different proposition. The manufacturers are aware they must offer some marketing punch to get people to hand over the money (if not technical punch). 16mp might be all that is necessary but Olympus said the same thing abut stopping at 12mp.

@Melvin, hasselblad also brought us an obscenely expensive rebranded sony nex camera in the past... so to me, that's not necessarily a reference.

@gwynoxy, correction, I believe that the mft consortium is on record for saying that 16mp is the sweet-spot for that given sensor size. it has nothing to do with what they think people want - it is a technical reason.

makes no sense because sensor shifting isn't really intended for handheld shots

Xamedes,

I've zero knowledge about Hasselblad camera, as well I do not know how the people use the Hasselblad pixel-shifting 200MP camera to take photos.

But, regarding to your "makes no sense because sensor shifting isn't really intended for handheld shots" for this rumored E-M5:

1. Per DP "If true, it is possible that Olympus will utilize its 5-axis stabilization system to achieve this"

2. As Photo 101, you need to disable the stabilization system while using tripod.

Make sense? or you need more explaining?

Actually ignore that last sentence.

I'd edit it out but can't see how to

I find it hard to wrap my head around how the extra pixels is going to help. A really good lens can resolve only 80 lines/mm. An m43 sensor at 16MP is already at 133 lines/mm (on the shorter axis).

Why not...

Dheorl, If this were true then cameras with good DR would have good ISO and vs versa in a predictable ratio 123 ISO = 1.23 DR but this is not the case. DR is not related to ISO. The d7100 at base ISO has DR exceeding cameras with better ISO performance.

Mosc, I am following that. But it is multiple s shots from the same sensor with the same exposure settings. This is not HDR, and so I see nothing which would improve base ISO or DR. Only potentially raise the pixel count.

Please read it again within context of this conversation and I am sure the DR comments will make sense. ISO performance does not improve by raising pixel count. Canon, for example, has multiple times raised pixel count without improvement to their ISO performance. That is a function of SNR not pixels. You can generally downsample to reduce noise but that is not the same as not having the noise.

Do you not know what SNR means?
Simply put it is amount of bad data to good data. Increasing the amount of data does not change this as it is a ratio. This is what is ISO noise. Read up on it at DXO or other site if you need more info.

deleted forum barfed and dup posted

That might mask the noise. I would be surprised if it did not. This is not the same as saying it has lower ISO noise. A d800 is 36mp, it has very low noise not because you can down sample the image but because the signal to noise ratio is very good.

Nope, you don't get it. I'm tired of explaining.

This is kind of like multi pass film scanning which primarily helps bring out details in the shadows. It really should help with DR and noise I think, but this technical stuff really isn't my strength. :)

You're right, Dheorl and Terry Breedlove, and Bjorn is just plain wrong. It's obvious that he doesn't understand how SNR works.

No Texinwien you are failing to understand something very very basic.
If you have an image which is 80% signal 20% noise. And you combine it with 3 more images which are the same ratio then the resulting image would still be 80% signal and 20% noise.... just 4 times as large.
Would the SNR be the same for 4 exposures? Yes, given that the exact same exposure settings, same scene and same sensor would be used.
Signal to noise for the thick headed or professional fan boys is just the amount of good data vs bad data in a data set. When you increase the data set size the ratio it contains is the same (DR & ISO is the same).
(80+80+80+80) = 320
(20+20+20+20) = 80
320+80= 400
320/400 = 80% signal
Down sampling can potentially hide some noise. But like I explained above this doesn't change how much noise was there originally.
The Nikon d800 has a good SNR because it has a higher % of good data to bad data, not because you can down-sample its 36mp images to some smaller image size.

Where that exact noise will be and how bad it will look is (as you say) random. However SNR is fairly predictable. So if you took 4 photos with the same settings on of the same scene with the same sensor there would be some variation in where the noise is and what it looks like but the actual amount of noise vs the amount of "noise-free" data would be the same.
Downsampling a 40mp image to a 16mp image will almost certainly produce a lower noise image. But as I said over and over, hiding noise by turning a 40mp image into a smaller size is not SNR it does not change how much noise there was only how much you see at a smaller resolution.
Consider the similar resolution Nikon d800. If it had a SNR of 90 at whatever ISO is was used at you can't change that image down to 16mp and claim it had an SNR of 99 simply because you managed to hide most of the noise by discarding data.
What you do in post is what you do in post. It has nothing to do with what the camera produced.

Wrong Bjorn. Totally wrong. SNR increases in quadrature. That is, noise increases at the square root of the increase in the number of photons captured. SNR = sqrt(N), where N on the right side equals number of photons captured when discussing photon slot noise. Of course, if we take 4 images of the same scene with the same settings, a blend of the four images will have the bedight of 4x as many photos as any of the single images, which will, in turn, give the blend a full so of improved SNR, compared to any of the single frames. I use this technique with excellent results where bracketing isn't practical. A blend of 16 photos of the same scene with the same settings quadruples SNR / increases it by two stops (compared to any of the single frames). Reference: http://en.wikipedia.org/wiki/Shot_noise , Examples: http://www.dpreview.com/forums/post/54799810

Correction: "a blend of the four images will have the bedight of 4x as many photos as any of the single images, which will, in turn, give the blend a full so of improved SNR" SHOULD BE "a blend of the four images will have the benefit of 4x as many photons as any of the single images, which will, in turn, give the blend a full stop of improved SNR". Stupid autocorrect on my tablet.

Still don't get the OBVIOUS texinwien?
You are ignoring the point that more photons aren't being captured, just the same amount over and over.
And more images aren't being blended. Blended is photo stacking which does not increase resolution, it would be a 16mp final image.
You are confusing HDR and similar techniques with photo stitching.

Sorry, Bjorn, but you're just completely confused about how SNR works. I suggest you take it to the Photographic Science and Technology forum, where you may end up learning a few things.

If we capture X photons in one exposure, we will capture 4X photons in four exposures. If we stitch or blend these together, the SNR of the resulting image will be 1 stop better than any of the single exposures. That's simple physics.

I'm not confusing this with HDR, either, since I do both HDR, blending (of equal as opposed to bracketed exposures) and stitching.

The bottom line is that SNR goes up with the number of photons captured, and blending or stitching multiple exposures obviously and logically means that the final image will have the benefit of having been created from more photons than any single capture, which, by the laws of physics means the final image will have a higher SNR than any of the individual exposures.

I suggest you read up on the effects of photo stitching vs HDR

I need to correct my previous post. "If we stitch or blend" should read "If we stitch and downsample or blend", and "and blending or stitching multiple exposures" should read "and blending or stitching and downsampling multiple exposures". I thought this would be clear from the context of the previous discussion, but re-reading suggests that it may not be.

The point is, if we combine multiple exposures, whether via stitching and downsizing or by blending, into a final image that is the same size as the original, individual captures, the resulting image will have higher SNR than any of the single captures.

That's just how SNR (due to photon shot noise) works.

What happens in POST PROCESSING, for example downsizing, has nothing to do with ISO or DR in the original image.

Indeed, it does - downsizing increases DR and SNR. Again, you're simply mistaken, and you don't know what you're talking about, or how SNR is calculated. I linked it above. Go read it and spend some time trying to understand it.

The simple rule is that the SNR of an image due to photon shot noise is sqrt(N), where N is the number of photons captured.

Can't argue with the laws of physics...

I'm not arguing physics, were I to do so you wouldn't understand that either. Your ignorant blather that downsizing has anything to do with photons is so far that it is mind boggling that anyone, no matter how ill-versed would post it.

I mean seriously, you think ANYONE is stupid enough to buy your claims that what you do in post-processing impacts the original image as captured? Good lord!

Glad to know you agree and admit that this m4/3 40mp is all a gimmick. So nothing real significant that olympus offer here.

like I said, it's use is nothing different than doing an HDR picture. just look at where and what the application is by Hasselblad and Olympus of the said sensor shifting. that sort of thing has been done in the past. sensor shifting to improve resolution of still, mostly inanimate objects in science, medical and landscape. different old tech but similar approach and methodology or concept in attaining high resolution which is very much like HDR. if rumor is true, then it is disappointing. it is still 16MP and would certainly not replace a higher MP camera's resolving capability no matter how you read and believe that 40MP marketing slogan. do that with a 20/24MP sensor + HDR, the higher MP camera would still produce better than that marketed 40MP resolution. there is just more pixels in higher MP cameras. there is no magic formula that a lower MP camera would produce better resolution than a native higher MP camera. only stupid would believe that.

take note that I use Sigma Merrills and make prints, and believe me that Merrills do not resolve that marketed 45MP resolution that is being said. it's in the math. all you need is use logic.

you can cry all you want daddyO. that 40MP marketing slogan is nothing more than a trick not different from doing an HDR picture. so don't be mad if someone is not blind as you to believe that it's the magical unicorn.

indeed!

at fmj1974, you have just described the EM1, not the EM5.

4K 30fps = 248 million pixels a second
1080p 60fps = 124 million pixels a second
720p 120 fps = 111 million pixels a second
40mp at even 5 fps = 200 million pixels a second

Yes, 40mp at considerable burst speeds is a serious amount of pixels to process ignoring the fact that the 40MP is a processing nightmare because of the math involved.

Really, 8 frames of 16mp even if the output is 40mp is basically worse than 128 mp (16x8) straight for the camera to process because it has to do more work. I doubt seriously that the EM5II would give more than 2fps in this mode, probably 1fps.

I think that Olympus and Sony are strong in technological innovations, other brands (mostly Panasonic, Samsung and Fuji) are also innovative, but mostly on camera features.

@SteB - methinks you are a bit one eyed, But - of course - I should really not be so cruel to take you out of your fluffy and rose Olympus fan dream :) There is lots of invention in digital cameras - and most are not made by Olympus. But - it does not really matter - does it?

OTF flash metering was invented and patented by Minolta. Olympus used it under license ( before Minolta used it themselves).
You forgot the f280 flash, wasn't that the first electronic flash you could use faster than the sync speed?

Olympus might seem like the most innovative camera company... if you're TWELVE. :o

Photography in it's core elements evolves very slowly. There are some key improvements which really changed it.
The idea of photography itself, flexible film rolls, automatic metering, color photography, digital sensor, AF, single lens cameras, image stabilization - these are some of the true milestones (of various importance, obviously) in the industry. Today we take them for granted.

Is anything Olympus did more important than the invention of digital sensor itself or implementing autofocus?
You've ever heard of Kodak or Minolta?

Seriously. If Olympus had suddenly disappeared with all its "inventions", photography would hardly change...

Who innovates most is such a forum peeing contest. What ever innovation turns out to be golden is good on its own merit however and the Super Sonic Wave Filter that came out in 2004 solved the dust problem conclusively before everyone else really nailed it (or tried to address the issue as the case may be). That was a good innovation. So is 5-axis IS. I'd say that Sony has good sensor innovation and their RX line and "7" line is quite innovative.

I agree that it sounds like a gimmick, but a similar technology has been used on recent high-end home theater projectors to shift the panels to display a higher resolution image. And from the reports I've read it actually works, allowing a 1080p projector to display an image that approached 4K resolution and detail.

So I'm excited to see it on a camera, and I don't think they would be doing it if it didn't work.

I admire Panasonic and Olympus for picking their sandbox and sticking to it. I am sure great innovation will come from trying to do more within the limits of the 4/3 although I think its approaching stationary phase of growth at this point. Definitely they make cameras that keep their value. An em1 is just as great a camera today as it was 2 y ago when it came out.

prossi,

The E-M1 has only been out for 1 year, not 2. But I agree that mine are just as great now, as they were a year ago.