PLEASE bring back eye-control AF!

pwm2 wrote in post #14860697
Unless I have read different documents than what you seem to be debating, the goal isn't to use the eye for autofocus, but to be able to replicate what our eye/brain does.

And then, it has nothing to do with the eye measuring any distance to the subject. What have been researched is figuring out how our brain performs image analysis to compute the likely distance to a subject without our eyes having to hunt around changing the focus distance.

So a camera implementation would then use the direction we look to figure out what to focus on, and use the image sensor together with advanced image processing to compute how much the lens should focus forward/backward from current position. So it would be similar to the contrast detection used in P&S, but instead of having the system perform many intermediate focus steps in front of and behind the true target, it should be able to (similar to the phase detection) compute how much off the lens is and send an optimal command to the lens.

You would then get a camera that wouldn't need a phase detect sensor but would use the image sensor instead - so just as a P&S camera, you could point at the image to specify where to focus. With your finger on a touch screen or with your eye in case the camera have eye-control. Or maybe with a cursor moved by a joystick.

The traditional contrast detection can only measure better/worse. It doesn't know how good the current solution is. And it doesn't know if it is behind or in front. So it always requires lots of iterations - and these iterations will fail if the subject moves. So in the end, there are many lens refocus operations and in-between there are waits to get out more readings from the sensor.

A phase-detect system can compute how much the lens is in front of, or behind, the subject, unless the focus is too much off or there are too much repeating features that gives false matches. So it basically perform two readings. One to decide how much to refocus. Then after the lens have been given that refocus command, the phase-detect system makes a second check to verify that all went ok. That allows the lens to use the maximum focus speed to make the full jump, and are the reason for the extremely fast focusing speed of a DSLR when fitted with a lens with a quick focus motor.

OK. It looks like I got my posts mixed up and lost something along the way. There was an article or a patent that used the method I described above where the focus is determined by the distance that they eye was focused using the flexing of the lens of the eye and I really thought I posted it but as I scroll back I can't find it. I guess when I get home I'll have to see if I can find it and post it. Sorry for the confusion. I must be getting old. lol

pwm2 wrote in post #14860697
Unless I have read different documents than what you seem to be debating, the goal isn't to use the eye for autofocus, but to be able to replicate what our eye/brain does.

And then, it has nothing to do with the eye measuring any distance to the subject. What have been researched is figuring out how our brain performs image analysis to compute the likely distance to a subject without our eyes having to hunt around changing the focus distance.

So a camera implementation would then use the direction we look to figure out what to focus on, and use the image sensor together with advanced image processing to compute how much the lens should focus forward/backward from current position. So it would be similar to the contrast detection used in P&S, but instead of having the system perform many intermediate focus steps in front of and behind the true target, it should be able to (similar to the phase detection) compute how much off the lens is and send an optimal command to the lens.

You would then get a camera that wouldn't need a phase detect sensor but would use the image sensor instead - so just as a P&S camera, you could point at the image to specify where to focus. With your finger on a touch screen or with your eye in case the camera have eye-control. Or maybe with a cursor moved by a joystick.

The traditional contrast detection can only measure better/worse. It doesn't know how good the current solution is. And it doesn't know if it is behind or in front. So it always requires lots of iterations - and these iterations will fail if the subject moves. So in the end, there are many lens refocus operations and in-between there are waits to get out more readings from the sensor.

A phase-detect system can compute how much the lens is in front of, or behind, the subject, unless the focus is too much off or there are too much repeating features that gives false matches. So it basically perform two readings. One to decide how much to refocus. Then after the lens have been given that refocus command, the phase-detect system makes a second check to verify that all went ok. That allows the lens to use the maximum focus speed to make the full jump, and are the reason for the extremely fast focusing speed of a DSLR when fitted with a lens with a quick focus motor.

Please see page 4 inclusive of paragraphs 20-25 of the patent, starting at "Summary of the Invention" where it is clearly stated that the system exclusively uses the distance determined by the eye to inform the camera where to focus and that the camera does not use the scene at all to focus.

kfreels wrote in post #14861757
OK. It looks like I got my posts mixed up and lost something along the way. There was an article or a patent that used the method I described above where the focus is determined by the distance that they eye was focused using the flexing of the lens of the eye and I really thought I posted it but as I scroll back I can't find it. I guess when I get home I'll have to see if I can find it and post it. Sorry for the confusion. I must be getting old. lol

K, this is the second article that you posted:

http://www.fastcompany​.com …ofocus-from-the-human-eye

This is more of a "This is how it works" style of article...It was in post #278.

The patent itself does cover this, and is more in line with this discussion.

apersson850 wrote in post #14860611
With an SLR camera of traditional design, the eye is focused on infinity all the time. You can see that if you open both your eyes. The viewfinder and the surroundings will be sharp at the same time.
Those who are near-sighted will need a correction lens, or use their glasses while shooting. Otherwise they can't see the viewfinder information properly.

Anders...I am not really sure of the exact focus distance of the eye when looking through the viewfinder(it would be great if our eyes had a distance scale).

I do know that, like JohnB57, my eyes will focus at infinity but due to age require (at least) +2 diopter in order to read. I thus must dial in +2 diopter for my viewfinder in order to be able to read SS aperture etc. on the bottom.

Where my eye is focused when looking at the scene through the viewfinder however, may be a different story and you may well be correct.

Regardless of whether the eye is focused at infinity or on the focus screen, either of these is a MAJOR obstacle to using the new system in a DSLR od EVF format.

As previously stated, I hope that research done since the patent was filed has solved this problem.

How does one know for certain that their eye is focused to infinity? Just curious. As I understand it, the eye focuses so fast that it is imperceptible.

I've long understood that the eye flicks from one point to the next and the next........ and the brain assembles these fragmentary and chaotic portions into some kind of order. But thats not what I understand Mark to be referring to.

I am 65 and need reading glasses. I see I've also set my camera viewfinder to +2 but unlike him I don't believe thats an insurmountable problem for a modern eye-control system.

kfreels wrote in post #14901492
How does one know for certain that their eye is focused to infinity? Just curious. As I understand it, the eye focuses so fast that it is imperceptible.

Lowner wrote in post #14901608
I've long understood that the eye flicks from one point to the next and the next........ and the brain assembles these fragmentary and chaotic portions into some kind of order. But thats not what I understand Mark to be referring to.

I am 65 and need reading glasses. I see I've also set my camera viewfinder to +2 but unlike him I don't believe thats an insurmountable problem for a modern eye-control system.

Hi guys...Maybe I should clarify my thinking in my previous posts, so that we will be on the same page...Or maybe I'm just on the wrong page(I hope so, as you know that I am a STRONG supporter of ECF) and you guys can let me know why.

I do think that focusing by using the distance determined by the eye is the way that ALL of autofocus is going to go, but I don't see how this system can be used in a single lens(both shooting and viewing) type of camera system, either reflex or EVF.

Here is the problem as I see it:

When viewing a scene through the viewfinder, the eye will be focused at one of three distances.
1.) On the focusing screen(my original contention) or closely focused
2.) Infinity(Apperson's contention) or far focused....Or
3.) Somewhere in between(perhaps even at the actual distance of the subject)

I'm really not sure which of these is correct, but they all three present problems when trying use this method for focusing a single lens camera system.

Case 1 is where the eye is close focused and clearly NOT focused at the correct distance to the subject and therefore would transmit to the camera the wrong distance to focus the lens.

Case 2 is where the eye is long focused(maybe "hyper-focally focused" would be a better term than "focused at infinity")...Focused like when you are driving rather than reading, where all but the closest objects are in focus all the way to the horizon.
Again the same problem as case 1 would exist if the subject happened to be close...The camera would receive the wrong distance information for critically focusing the lens.

Case 3 would apparently send the correct focusing distance to the camera, but I see a "chicken and egg" problem here...The eye would have to focus on the subject through an unfocused single lens in order to tell the camera the distance to focus the single lens. Perhaps not so bad if the lens is only slightly defocused, but what if the lens is grossly defocused so that it is hard for the eye to even find the subject, much less focus on it? And what effect would using 600mm as opposed to 16mm have on the focusing distance as perceived by the camera through the eye?

Maybe as so often happens, technology of which I am unaware has already solved these problems or soon will, but I don't have the answers to these questions and perhaps I am wrong or am overlooking something simple(wouldn't be the first time lol).

I'm open to suggestions...

Gee, Where'd everybody go?...I hope I didn't end the thread.

Mark,

I have no answers, but I found yesterday I was struggling to see the exposure correction scale in the viewfinder of my 5D2, so wound the dioptre adjustment as far in the plus direction as it would go and now its fine. But while doing so I carefully watched the scene above and it made not one jot of difference one way or the other. I'd always assumed it was focused onto a piece of ground glass (or plastic?) and that we focus on that. But my experience yesterday has confused me.

Interesting observation Richard, I hadn't thought to try something like that.

I'm getting ready to retire for the evening so will try that in the morning and see if I can add any insight.(I'm an oddball and stay up all hours...It just occured to me that you are probably having your morning tea, as I get ready for bed...Enjoy!)

Thanks for responding.

I highly doubt canon is listening. People wanted a rangefinder dslr but instead they released the "eos- M" system which is more of an s100 upgrade rather than a fuji x100 competitor. I still do think it would be a nifty idea to implement.

Canon is continuing to update this document (Eye controlled focus and focus techniques) about which digital SLRs have eye controlled focus. The 1Dx, EOS-M and T4i are now listed.

c2thew wrote in post #14934020
I highly doubt canon is listening. People wanted a rangefinder dslr but instead they released the "eos- M" system which is more of an s100 upgrade rather than a fuji x100 competitor. I still do think it would be a nifty idea to implement.

You may be right, but I hope you're wrong...Too bad about that EOS-M...Canon missed an opportunity there.

macroimage wrote in post #14935513
Canon is continuing to update this document (Eye controlled focus and focus techniques) about which digital SLRs have eye controlled focus. The 1Dx, EOS-M and T4i are now listed.

Very interesting...I wonder what that means?


By the way guys...Here is a "wish" thread, where you can vote for the feature that you would most like to see in a DSLR:


https://photography-on-the.net …/showthread.php​?t=1223174


We all know what that should be don't we?...Just a wish, but the more votes the better, huh?

Eye control taking a step into TV's now. http://www.engadget.co​m …r-eye-control-tv-eyes-on/
I've said it before and I'll say it again......eye control is the future and Canon was too early. Time to get this going again. Can we please have a 7D II with dual digic 5, 12 fps, 1.3X sensor, ISO 51,200 (clean native to 12,800) ECF as described in my previous posts and dual card slots? That would really make my day!

kfreels wrote in post #14950817
Eye control taking a step into TV's now. http://www.engadget.co​m …r-eye-control-tv-eyes-on/
I've said it before and I'll say it again......eye control is the future and Canon was too early. Time to get this going again. Can we please have a 7D II with dual digic 5, 12 fps, 1.3X sensor, ISO 51,200 (clean native to 12,800) ECF as described in my previous posts and dual card slots? That would really make my day!

I like this idea K...A 7DII as described above with the best possible ECF that can be managed with current technology...Obviously if they can do this with a TV being controlled by someone sitting across the room, it would be much easier using the eye at a viewfinder.

Maybe determining distance directly from the eye is too far advanced at the moment, but I'm willing to take whatever is the best that they can do for now.

Bring on the 7DIIE!...I'll buy one now...Canon are you listening?

And again. To keep from having to re-read this whole thread, the idea for the ECF is as follows......
It works exactly as it does now. You press the button to select your AF point. Only instead of having to move the scroll wheel, QCD or joystick to select your point, the ECF will track your eye along the AF points as it moves. Press the button again to make it stick. Any erroneous selection can be immediately dealt with via the normal QCD, scroll wheel and/or joystick just as it is now. They will remain active even while the ECF is working. Not that complicated really.