Dynamic range of the human eye

[AlanF]

There are ongoing discussions about the dynamic range of the human eye versus that of sensors so I asked a professor in the area of human vision for the latest ideas on the DR of light receptors in the eye.

We can see (something) at illuminance levels from about 10^-4 lux (dark night) up to 10^5 lux (full sunlight) so one might say the range is 10^9 (= 30 stops). But, it doesn’t use that range all at once. The key difference between the eye’s light receptors and a camera sensor is that the sensor has a linear response to light intensity whereas the light receptors in the eye do not. Individual receptors can alter their sensitivity by events surrounding them and the levels of light around the areas they are seeing. Sets of receptors can adapt to seeing light areas or dark areas simultaneously with different sensitivities, effectively like HDR in a camera or pushing shadows in post-processing. The bright light range is increased because in very bright light many or most, but not necessarily all, of the receptors become bleached so we can still see highlights from the ones that are not bleached.

The physiology is very interesting and it does give justification for lifting shadows by post-processing because it does mimic what we do in natural vision with our own eyes.

 

[NancyP]

I have the first edition of this book "Science for the Curious Photographer", Charles Johnson, and it is an excellent book.

 

[Maximilian]

...
The physiology is very interesting and it does give justification for lifting shadows by post-processing because it does mimic what we do in natural vision with our own eyes.

But only, if you are looking in the shadows only! Not, if you have bright and dark parts in your vision at the same time. Then the human eyes are not able to resolve both at the same time.

I have no problem with shadow lifting. I have no problem with any kind of PP adaption if somebody likes to do so. Anybody should do as he/she desires.

But if you have a high DR in a scene the eye cannot resolve both at the same time.
A HDR picture of that scene quite often looks artificial, if the HDR is not used very carefully.
One might like that look or not, that's taste and artistic interpretation.
But the reason that it looks artificial is, that your visual experience does not match that result.

So I cannot agree with your conclusion in this point.

 

[AcutancePhotography]

One of the problems is that the eye does not operate like a sensor on a camera. This makes such comparisons difficult. Not only is the eye constantly adjusting; at the same time, the brain is constantly performing its own version of image processing.

 

[privatebydesign]

Not having a friendly professor to hand I go to sites i trust.

The most consistent figures I have seen is >24 stops within a single scene but not without changing the pupil size, which we do imperceptibly.

As for the adjustable receptors, analogous to different iso ratings per pixel, that ability still only covers around 11-14 stops. 11-14 stops is, as I understand it, the range of luminance we can do without adjusting the pupil (aperture).

 

[Sharlin]

One must also remember that the human eye has two separate sensor systems: the less-sensitive but color-sensing cone cells and the more sensitive monochrome-only rod cells. In normal lighting conditions, the rod cells are constantly saturated and their signal is ignored. When it gets dark, it takes up to half an hour for the eye to fully adjust because the photopigments in the rods need to chemically reset to their non-excited configuration.

 

[Sharlin]

...
The physiology is very interesting and it does give justification for lifting shadows by post-processing because it does mimic what we do in natural vision with our own eyes.

But only, if you are looking in the shadows only! Not, if you have bright and dark parts in your vision at the same time. Then the human eyes are not able to resolve both at the same time.

Plain eyes, no. But what we actually perceive is far removed from the raw sensory data coming from the eyes. Among all sorts of creative interpretation that the brain does, it integrates visual imagery over time to form a stable, coherent, high dynamic range mental image.

 

[docsmith]

Thanks for the research and post. I have seen the classic range of typical lighting from +16 EV (bright sun) to -6 EV (moonless night)....so had always assumed it was ~22-24 stops. Because of that, I have always assumed the ultimate goal would be to approach this range as that would capture and represent pretty much everything we can perceive in nature in a single image.

The fact that it is 30 stops is interesting. My guess would be that we can "see" brighter than a bright day. This could become interesting in some artificially lit areas.

 

[privatebydesign]

Thanks for the research and post. I have seen the classic range of typical lighting from +16 EV (bright sun) to -6 EV (moonless night)....so had always assumed it was ~22-24 stops. Because of that, I have always assumed the ultimate goal would be to approach this range as that would capture and represent pretty much everything we can perceive in nature in a single image.

The fact that it is 30 stops is interesting. My guess would be that we can "see" brighter than a bright day. This could become interesting in some artificially lit areas.

Well if you change the aperture and shutter speed on a camera it has infinite DR. The real point of interest is the constant pupil range, that is the same as a single exposure, and it turns out our eyes and cameras are very close in that measure.

 

[Loswr]

Thanks for the research and post. I have seen the classic range of typical lighting from +16 EV (bright sun) to -6 EV (moonless night)....so had always assumed it was ~22-24 stops. Because of that, I have always assumed the ultimate goal would be to approach this range as that would capture and represent pretty much everything we can perceive in nature in a single image.

The fact that it is 30 stops is interesting. My guess would be that we can "see" brighter than a bright day. This could become interesting in some artificially lit areas.

Well if you change the aperture and shutter speed on a camera it has infinite DR. The real point of interest is the constant pupil range, that is the same as a single exposure, and it turns out our eyes and cameras are very close in that measure.

Except that the eye can change aperture and 'shutter speed' much, much faster than someone holding a camera. Plus, the eye has a wicked good exposure meter.

 

[Sharlin]

Thanks for the research and post. I have seen the classic range of typical lighting from +16 EV (bright sun) to -6 EV (moonless night)....so had always assumed it was ~22-24 stops. Because of that, I have always assumed the ultimate goal would be to approach this range as that would capture and represent pretty much everything we can perceive in nature in a single image.

But the human eye (plus brain) isn't even remotely capable of seeing >20 stops in a single scene... so why should that be the goal for camera technology? Plus compressing 20 stops of DR into <10 stops for display would be a fool's errand anyway...

 

[privatebydesign]

Thanks for the research and post. I have seen the classic range of typical lighting from +16 EV (bright sun) to -6 EV (moonless night)....so had always assumed it was ~22-24 stops. Because of that, I have always assumed the ultimate goal would be to approach this range as that would capture and represent pretty much everything we can perceive in nature in a single image.

The fact that it is 30 stops is interesting. My guess would be that we can "see" brighter than a bright day. This could become interesting in some artificially lit areas.

Well if you change the aperture and shutter speed on a camera it has infinite DR. The real point of interest is the constant pupil range, that is the same as a single exposure, and it turns out our eyes and cameras are very close in that measure.

Except that the eye can change aperture and 'shutter speed' much, much faster than someone holding a camera. Plus, the eye has a wicked good exposure meter.

As well as the better exposure meter the processor is pretty advanced too, in Canon terms evolution of the eye brain processing is probably on to DiGiC 1,376,598.

 

[raptor3x]

Not having a friendly professor to hand I go to sites i trust.

The most consistent figures I have seen is >24 stops within a single scene but not without changing the pupil size, which we do imperceptibly.

As for the adjustable receptors, analogous to different iso ratings per pixel, that ability still only covers around 11-14 stops. 11-14 stops is, as I understand it, the range of luminance we can do without adjusting the pupil (aperture).

Is this based on that Cambridge in Colour article? If so, you should know that article is mostly nonsense, the eye isn't rapidly changing aperture and then combining exposures in the same way that we do by combining multiple shots at different exposures. Like the original post stated, the large dynamic range comes from local sensitivity adaptation of the retina and you do indeed get a much larger dynamic range in one "exposure" than what any modern camera can achieve.

 

[privatebydesign]

Not having a friendly professor to hand I go to sites i trust.

The most consistent figures I have seen is >24 stops within a single scene but not without changing the pupil size, which we do imperceptibly.

As for the adjustable receptors, analogous to different iso ratings per pixel, that ability still only covers around 11-14 stops. 11-14 stops is, as I understand it, the range of luminance we can do without adjusting the pupil (aperture).

Is this based on that Cambridge in Colour article? If so, you should know that article is mostly nonsense, the eye isn't rapidly changing aperture and then combining exposures in the same way that we do by combining multiple shots at different exposures. Like the original post stated, the large dynamic range comes from local sensitivity adaptation of the retina and you do indeed get a much larger dynamic range in one "exposure" than what any modern camera can achieve.

Where does it put a figure on that in humans?

 

[docsmith]

Plus compressing 20 stops of DR into <10 stops for display would be a fool's errand anyway...

I'd agree with this wholeheartedly and have had this debate a couple of times. Jpgs are 8 stops, most monitors are 8 stops.

But, and this is the counter argument I do agree with, the human eye/brain can adjust so quickly, so quickly that a very wide dynamic range is compressed in how the viewer remembers the scene. Sometimes, you are trying to recreate that in a single static image. That's when pulling from a wide DR scene can look natural when compressed into a 8 stop jpg. So, maybe we do not need 24 stops of DR in a camera, but I would argue we also need more than 8 stops. Of course, the 12-14 stops that most current cameras covers the vast majority of what is needed. I can see going a bit higher, but it is diminishing returns.

 

[Mt Spokane Photography]

There are ongoing discussions about the dynamic range of the human eye versus that of sensors so I asked a professor in the area of human vision for the latest ideas on the DR of light receptors in the eye.

We can see (something) at illuminance levels from about 10^-4 lux (dark night) up to 10^5 lux (full sunlight) so one might say the range is 10^9 (= 30 stops). But, it doesn’t use that range all at once. The key difference between the eye’s light receptors and a camera sensor is that the sensor has a linear response to light intensity whereas the light receptors in the eye do not. Individual receptors can alter their sensitivity by events surrounding them and the levels of light around the areas they are seeing. Sets of receptors can adapt to seeing light areas or dark areas simultaneously with different sensitivities, effectively like HDR in a camera or pushing shadows in post-processing. The bright light range is increased because in very bright light many or most, but not necessarily all, of the receptors become bleached so we can still see highlights from the ones that are not bleached.

The physiology is very interesting and it does give justification for lifting shadows by post-processing because it does mimic what we do in natural vision with our own eyes.

Are we talking Dynamic Range, or ISO equivalence? I think the term is being used more like ISO equivalence, where as in photography, we use dynamic range for a sene that happens in a short period.

 

[ethanz]

one might say the range is 10^9 (= 30 stops).

Why can't Canon make 30 stops of DR? God and Sony can. Time to ditch Canon.

;D ;D :

 

[edoorn]

Wait - my brain creates jpgs out of my eyes? I want raw!!

 

[Jack Douglas]

Wait - my brain creates jpgs out of my eyes? I want raw!!

I want an eye transplant with Hooty as the donor. Think how cute I'd be and spying would be a breeze - they do have colour vision don't they?

Jack