Author Topic: New camera that doesn't over-expose  (Read 4713 times)

ItsMeLenny

  • Hero Member
  • *****
  • Posts: 928
  • 550D
New camera that doesn't over-expose
« on: September 01, 2015, 03:03:28 PM »
http://petapixel.com/2015/08/18/mit-created-a-camera-that-will-never-overexpose-a-photograph/#more-176916

This is a new camera that will allegedly never over expose a photograph.
It works like water raining into a cup :P
But then the water auto levels because of the fluidity of rain and man combined.

dfort

  • Guest
Re: New camera that doesn't over-expose
« Reply #1 on: September 01, 2015, 04:40:21 PM »
Well this illustration explains everything:



Seriously though it sounds a bit like the old "expose for the shadows, develop for the highlights" mantra when I was learning photography.


dmilligan

  • Developer
  • Hero Member
  • *****
  • Posts: 3218
  • 60Da / 1100D / EOSM
Re: New camera that doesn't over-expose
« Reply #2 on: September 01, 2015, 05:34:14 PM »
To extend your analogy, from what I understand from the article: when the cup fills, you dump it out and then keep a count of how many times you have to dump each cup out. This means you can capture huge dynamic range, however it doesn't make it actually impossible to overexpose a photograph. Because of course you will have to use a digital number to keep track of your cup overflow count, and of course this number's precision will have to be limited to some finite number of bits, which means your cup overflow count can itself overflow, which means you could in face overexpose an image. Considering that you need to keep a separate count for each cup (pixel), you will either need a very huge number of bits to store all of these overflow counts (very expensive hardware), or you will have to limit the bits to a relatively small number per cup, which makes overflowing your overflow count more likely. Ultimately your DR is still limited by the total number of bits you use to represent each pixel, because digital values themselves will always be represented in a finite manner and have the potential to overflow.

So I believe the title of this article is misleading and somewhat clickbait. It should really read: MIT created a camera that can capture very large dynamic range in a single exposure.

vyskocil

  • Freshman
  • **
  • Posts: 53
Re: New camera that doesn't over-expose
« Reply #3 on: September 01, 2015, 08:19:31 PM »
As I understand it the counts each cup has been flushed is not stored, the image is reconstructed from the modulo image with some stochastic algorithm as explained in the video (markov chains, etc...)

Audionut

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 3657
  • Blunt and to the point
Re: New camera that doesn't over-expose
« Reply #4 on: September 02, 2015, 04:46:20 AM »
http://web.media.mit.edu/~hangzhao/papers/moduloUHDR.pdf

Quote
To design an imaging sensor that has infinite dynamic range is physically infeasible, since the sensor keeps on collecting photons but the storage and precision of analog-to-digital converter (ADC) cannot increase infinitely. A smart tradeoff in taking ultra high dynamic range data with a limited bit depth is to wrap the data in a periodical manner. This creates  a  sensor  that  never  saturates:  whenever  the  pixel value gets to its maximum capacity during photon collection,  the  saturated  pixel  counter  is  reset  to  zero  at  once, and  following  photons  will  cause  another  round  of  pixel value  increase.   This  rollover  in  intensity  is  a  close  analogy to phase wrapping in optics, so we borrow the words “(un)wrap” from optics to describe the similar process in the intensity domain. Based on this principle, a modulo camera could be designed to record modulus images that theoretically have an Unbounded High Dynamic Range (UHDR).

In  this  paper,  we  explore  the  use  of  a  modulo  camera in both single-shot and multi-shot scenarios to address the HDR  problem  from  a  brand  new  point  of  view.   To  extend  the  dynamic  range  with  a  single  shot,  we  propose  a graph-cuts-based unwrapping algorithm to recover information in the wrapped region (saturated in the intensity image). The  unwrapped  result  from  an  8-bit  modulus  image  has much higher dynamic range than a single image captured by a conventional 8-bit intensity camera

I haven't read much yet, but it sounds like they keep track of the number of times the pixel well gets emptied.  That makes sense too.

however it doesn't make it actually impossible to overexpose a photograph. Because of course you will have to use a digital number to keep track of your cup overflow count, and of course this number's precision will have to be limited to some finite number of bits, which means your cup overflow count can itself overflow, which means you could in face overexpose an image. Considering that you need to keep a separate count for each cup (pixel), you will either need a very huge number of bits to store all of these overflow counts (very expensive hardware), or you will have to limit the bits to a relatively small number per cup, which makes overflowing your overflow count more likely.

If you're exposing the sun for 30 seconds, yeah I guess so.  I don't expect the pixel wells need to be emptied a large number of times in most use cases.  Your point does stand though.


Ultimately your DR is still limited by the total number of bits you use to represent each pixel, because digital values themselves will always be represented in a finite manner and have the potential to overflow.

For this camera, I thinks it's more accurate to say that the DR is limited by the total number of bits you use to represent each pixel well empty.  Since pixel well capacity is a trade off, you could theoretically use small pixel wells, combined with a high capacity to record pixel well empties.  All the benefits of smaller pixel wells, combined with the saturation capacity of larger pixel wells*:)

So I believe the title of this article is misleading and somewhat clickbait.

Aren't they all these days.  ::)

*Minus shot noise benefits.

Seriously though it sounds a bit like the old "expose for the shadows, develop for the highlights" mantra when I was learning photography.

Exactly.  You don't gain the shot noise benefits from capturing all of those "highlight" photons (since you throw photons away every time you empty the pixel well), but this is splitting a fine hair.

Actually, the more I think about it, if the pixel well only just gets emptied, then the exposure stops...............