Author Topic: How about 250Mpixels?  (Read 8216 times)

Peter Forsell

  • NG Member
  • *
  • Posts: 425
  • A Cunning Linguist
Re: How about 250Mpixels?
« Reply #15 on: September 09, 2015, 13:48:10 »
If (that's still a big if) the noise characteristics of higher resolution sensors do not get worse than what's currently available, there is still a lot of room for more pixels. The systemic MTF depends both on lens resolution and sensor resolution. Put a higher res sensor behind a lens and you'll get a higher resolution image, and vice versa. The limit for FF sensors is still pretty far ahead, about 14 gigapixels for 550 nm light, so 250 or 500 megapixels is not unreasonable. Higher resolution will solve aliasing and moiré issues perfectly when the lens acts as a low pass filter. And we have the option of downsampling, of course. Let the camera capture at high resolution and we can downsample or decimate 1:4 or 1:8 or 1:16 as required. As we all know, a downsampled 12 megapixel image from a 36 megapixel camera looks better than a native 12 mexapixel image.

We are on a curve of diminishing returns. Next step in sensor resolution should be double the resolution of today. Anything less won't really be visible.

In the big picture higher resolution has meant lower noise. Every time. Take Nikon D1X -> D2X -> D3X -> D810 -> ...
Or take D1 -> D1H -> D2H (dubious?) -> D3 -> D4 -> ...

 So, how about a 64 megapixel D900, then 128 megapixel D950 after a few iterations of oil, dust, flare, left-focus-issue (D910, D920, D930...).


Jan Anne

  • Noob
  • Global Moderator
  • **
  • Posts: 2045
  • Holland
    • Me on Flickr
Re: How about 250Mpixels?
« Reply #16 on: September 09, 2015, 15:09:41 »
While Canons 250MP sensor is not made for DSLR use they did announce yesterday that they are working on a 5D like 120MP DSLR:
http://www.canon.com/news/2015/sep08e2.html
Cheers,
Jan Anne

Akira

  • Homo jezoensis
  • NG Supporter
  • **
  • Posts: 12823
  • Tokyo, Japan
Re: How about 250Mpixels?
« Reply #17 on: September 09, 2015, 17:18:31 »
Not at all Akira. I can handhold a 6 gigapixel camera too.
...avoiding blur will be a major challenge though  ;D ;D ;D

Well, Sten, a 6GP camera might be heavy enough to absorb the mirror and shutter slap as well as the hand blur!
"The eye is blind if the mind is absent." - Confucius

"Limitation is inspiration." - Akira

Peter Forsell

  • NG Member
  • *
  • Posts: 425
  • A Cunning Linguist
Re: How about 250Mpixels?
« Reply #18 on: September 11, 2015, 10:47:44 »
Well, Sten, a 6GP camera might be heavy enough to absorb the mirror and shutter slap as well as the hand blur!

How much does a pixel weigh? Is D750 twice as heavy as D700?  :P

Akira

  • Homo jezoensis
  • NG Supporter
  • **
  • Posts: 12823
  • Tokyo, Japan
Re: How about 250Mpixels?
« Reply #19 on: September 12, 2015, 07:26:59 »
How much does a pixel weigh? Is D750 twice as heavy as D700?  :P

Peter, the 12MP D700 even without the lens is nearly three times heavier than 4MP Coolpix4500!
"The eye is blind if the mind is absent." - Confucius

"Limitation is inspiration." - Akira

Andy

  • NG Member
  • *
  • Posts: 314
Re: How about 250Mpixels?
« Reply #20 on: January 31, 2016, 15:44:10 »
Having 19580x12600 in a 29.2x20.2mm, great! Considering the wavelength of visible light, that's about twice the minimum theoretical size of a photosite, assuming perfect optics, of course. It might have applications for research where they sometimes use electronic microscopes to "see" things too small for regular light waves.

Sorry if I joined this interesting thread too late :-)

Personally, I consider this a good step in the right direction, but there is still a long way to go, to be of transformative value to photography.

Just 3 quick thoughts:

1) Current sensor technology is in a large part still analog technology (until the A/D process kicks in) with the known broad set of noise sources. Imagine that the size of the sensels will get over time so small that each sensel can detect/register a single photon hitting the sensor surface - effectively moving the digital domain much further upstream. No need to check for full well capacity, no need for perfect calibration among the many A/D converters, etc, etc, ..). Removes some classes of noise sources.

2) It allows to use software to address current physical limitations. Take diffraction limits: Well know, often an issue, etc ... Imagine you have a sensor with such a resolution, that you have 10000 sensel per Airy disc. Allowing the camera to measure the shape, amplitude and size of the Airy disc. If there would be only one Airy disc, processing would be simple. Due to overlapping nature of all the Airy discs in a photographic image, an inference pattern of overlapping airy discs will be the result. With sufficient specialized and parallel CPU resources, the image degradiation of diffraction can potentially be computationally compensated (partially or may be even fully).

A fictous example:
Internal resolution: 10 Gigapixel on DX
External resolution: 100 MP for the RAW file
Wouldn't it be nice to use your (from a user's perspective) 100 MP camera even at f16 without beeing impacted by diffraction ? :)
The current technology approach wouldn't get you there.

3) I assume, that sooner or later, the notion of RAW or sensor resolution will split into "2 resolutions". One internal to the camera and one "visible" to the photographer in selecting the RAW size option of the camera. The native chip resolution will not be a 1:1 relation to the storage resolution like today. Kind of oversampling. BTW, Nikon's D journey started along those lines. The orginal D1 sensor used 4-fold binning internally. Only option to store the RAW was the 2,7 MP resolution (but not the "native" 10,8 MP of the sensor).

This decoupling might open up another option for camera manufacturers. It allows them to have different speed of progress for the "native&internal" resolution increases HW technology will provide and the "storage" resolution they will be able offer to the user (depending on the sppeed of progress in SW, image processing, etc, etc, ...)


Just thoughts,
Andy


Hugh_3170

  • NG Supporter
  • **
  • Posts: 2127
  • Back in Melbourne!
Re: How about 250Mpixels?
« Reply #21 on: January 31, 2016, 15:56:54 »
Andy, if we double the number of pixels, then the resolution will only go up by around 41.4%.

Conversely if we wish to double the (linear) resolution, then we would need to quadruple the pixel count of the sensor.  Thus, for example, a 12MPx camera has only twice the linear resolution of a 3Mpx camera, and not four times.
Hugh Gunn

Andy

  • NG Member
  • *
  • Posts: 314
Re: How about 250Mpixels?
« Reply #22 on: January 31, 2016, 16:27:59 »
Hugh,
I agree with you on your statement, but I am not sure how it relates to my post above.

My intention was to indicate potential benefits of cameras with > 1 Gigapixel on the level of the sensor without surfacing this "native" resolution to the user for further handling and processing, like most contemporary cameras do.

And if the linear or area resolution goes up 2x, 3x, 4x, 9x, 16x just reflects the steplevel changes of possible technical and economic improvements in this longer "journey" (if it happens at all).
 
rgds,
Andy 

Ilkka Nissilä

  • NG Member
  • *
  • Posts: 1712
  • You ARE NikonGear
Re: How about 250Mpixels?
« Reply #23 on: January 31, 2016, 17:37:06 »
While photon counting has its advantages, they would only show up in the deepest of the shadows and be meaningless to photography - that deep shadows would still be noisy even if you use photon counting. And the cost in terms of computation and electrical power would likely be prohibitative.

You cannot undo diffraction; it is a lossy process. You can apply a  deconvolution operation tailored to counter the effects of diffraction with a particular lens and f-stop to make the image sharper but like all sharpening operations, it increases the contrast of noise. You are not gaining any new information that way. Deconvolution based sharpening operations have been available in image editing software for over a decade.

Andy

  • NG Member
  • *
  • Posts: 314
Re: How about 250Mpixels?
« Reply #24 on: January 31, 2016, 18:33:23 »
You cannot undo diffraction
I am not an expert in this matter, but there seems to be ample research going on in the field of sub-diffraction-limited (SDL) imaging.

Eric Fossum, the inventor of the CMOS sensor has written a paper arguing for further work in such a direction, and if I remember correctly, SDL research papers were published in Nature as well.

rgds,
Andy


Ilkka Nissilä

  • NG Member
  • *
  • Posts: 1712
  • You ARE NikonGear
Re: How about 250Mpixels?
« Reply #25 on: January 31, 2016, 22:53:07 »
Fluorescence microscopy is quite different from photography. I don't see anything in those methods (e.g in Nature Photonics) that could be applied to general photography.

As for Fossum's idea of using extremely small pixels, it may allow recovery of finer details if there is unlimited light (to avoid photon noise from blowing up).

Tristin

  • NG Member
  • *
  • Posts: 1083
  • Nothing less, always more.
Re: How about 250Mpixels?
« Reply #26 on: February 01, 2016, 04:56:10 »
Even with the prospects of reducing moire, diffraction, etc., I still feel there are bigger issues to tackle.  Nikon and Canon are losing plenty of customers to mirrorless, and it is not for anything that sensor tech will improve.  I'd be infinitely more interested in tech that combined the advantages of EVF and mirrorless to DSLRs and OVF.  Whether it is an effective hybrid VF or a hybrid of a mirrorless/mirrored camera, that would improve photos far more than images getting a litter crisper and having less moire.

As for Canon's megapixel pursuit, they have a broader scope of customers they sell to than Nikon, so I'm sure what they choose to develop is not weighed soley on the still photography side of their business.  With video (a big business for them) continues to dominate at a rapid pace, I would not surprised if their push to stay competitive in stills lags further.  Meaning there will be tech trickled to their photography department, but I doubt that there photography products will be getting the emphasis.
-Tristin

Ilkka Nissilä

  • NG Member
  • *
  • Posts: 1712
  • You ARE NikonGear
Re: How about 250Mpixels?
« Reply #27 on: February 01, 2016, 13:09:25 »
I am not sure how big the video market is for Canon either, but surely they are taking it very seriously. They have developed the dual-pixel PDAF system which seems a very good way to focus during live view and video, and as processing power increases, it will get progressively more sophisticated. AFAIK the C300 II can do fluid continuous focusing during video based on virtually any area of the frame. But it costs $16k. Not exactly a camera for the masses. Dual pixel AF with continuous focusing will likely eventually be offered on less expensive models and still cameras, including mirrorless. If I'm not mistaken, in many less expensive models where dual pixel AF is offered it doesn't track moving subjects yet.

Consumer video interest comes and goes, in principle people are interested in it but few people do proper sound recording and editing which means most of the material is a suffering to watch. I guess stabilization helps a little to achieve video than is pleasant to watch with less effort than in the past. On the other hand our subconscious references are in Hollywood movies and there the technical quality is a bit higher than what one can achieve with a crew of one. Thus people may get excited about video and then disappointed because of the huge amount of work that is needed to make it really good. External audio recording makes a huge difference to the quality of the whole viewing experience but requires extra gadgets. I mainly see video as a professional feature, not so much for consumers (yes, I know there is a lot of stuff on youtube, vimeo etc. but ...) Still photography by contrast is relatively easy for an individual without crew to accomplish to a high standard, thus it is popular and there is a wider user base.  In still photography Canon is still very popular because of their extensive lens lineup. And the low ISO dynamic range is really of interest to a minority - the majority see Canon's image quality as good enough and focus on content, or at least I'd hope so.

Mirrorless has gotten traction and those system cameras are now very common to see on the streets.  I think it is a good thing that there are different technological approaches for customers to choose from, with different advantages and disadvantages, for different users and applications. I don't see a reason why every camera manufacturer should have to have a presence in every part of the camera market. It becomes very expensive to develop a product for every category and try to make it better than everyone else's products in those categories. Mirrorless and DSLR use different focusing technologies and thus unless the user wants to use manual focus,  there are problems in cross-platform application of autofocus lenses (which would still exist within the brand's own products, e.g. Nikon 1 series AFs only using the center focusing point, and tracking isn't good, when mounting F mount lenses). Sony with adapters allow varying degrees of autofocus compatibility with Canon lenses but would you not want optimal autofocus performance with every lens in the system? If it is not possible to have everything working like it does on DSLR or better then why bother?  In reality it is likely possible for a manufacturer to be the best in a few product categories and then the question is what is the purpose of those cameras which are not competitive?  I think it is better that each company focuses on one or two product lines and takes their development to the hilt of what is possible, and then the different technologies with optimal implementations can compete with each other. If a company does a bit of everything then likely none of it will be the best.

At work I use GH4R with Leica 25/1.4 for video (and a Zoom recorder for audio), and it is very nice. We do recordings of measurement sessions that take place in a dim room and the low light video quality of that camera is excellent. It is easy to focus also, with the magnified view that pops up as one turns the focus ring. I don't think with Nikon it would be as nice as the manual focus rings of most AF(-S) Nikkors are poor. So why even bother going there if they're not going to do the MF right. Of course, there are a lot of manual focus lenses available for Nikon, but I would like to use modern Nikon lenses rather than old ones or ones made by Zeiss. Other manufacturers have shown it is perfectly possible to have precise MF in an AF lens. So I would choose the system for the application instead of trying to use a system that was created for something different for a purpose where it is an ill fit. The micro four thirds system fits well for our photogrammetry as well as video recordings of our sessions and Nikons would not. However, for my photography (outside of my research work) I prefer DSLRs and their optical viewfinders.

simsurace

  • NG Member
  • *
  • Posts: 835
Re: How about 250Mpixels?
« Reply #28 on: February 01, 2016, 18:10:53 »
Getting more photons to reach the photosite and getting the photons to trigger more electrons has a higher benefit than counting individual photons.
Simone Carlo Surace
suracephoto.com