[pixelninja]

A student in the Computer Science department at Stanford has been working on a camera that samples the full 4D light field inside the camera in a single photographic exposure. Basically what this means is that with this camera, you can take a single picture, and then adjust the focus after the exposure has been taken!

http://graphics.stanford.edu/~renng/

Ding! My noddle has officially been baked...

 

[BAH]

its amazing to me how some peoples brains work. I sometimes can't figure out how to open tool packaging without 4 knives, a pair of scissors and a firebomb

 

[splat]

I see where he is going and it is very interesting . I just don't agree that it is a 4 Dimensional Field he is capturing. It sounds to me like he is capturing the entire 3 D world and then you can get to any 2d plane to have the picture you want . quite an interesting concept.

 

[stinkyboy]

Here's what I don't understand: "I was also the president of the Stanford Climbing Wall for a year"

 

[BigMike]

Wow. That is amazing!

If this takes off, that dude is gonna be RICH!

 

[SK6]

I see where he is going and it is very interesting . I just don't agree that it is a 4 Dimensional Field he is capturing. It sounds to me like he is capturing the entire 3 D world and then you can get to any 2d plane to have the picture you want . quite an interesting concept.

Thats EXACTLY what I was thinking.....more of a 3d focal adjustment per say.

 

[Discostu]

Wow. That is amazing!

If this takes off, that dude is gonna be RICH!

I'm sure stanford will get most of the money since he is using their equipment and resources to do research. It sucks, but thats the way it goes. At my school everyone says if you have a good idea to wait until you graduate to do anything with it. There are a few student projects that have become businesses recently and all the profits are going to the University.

 

[SK6]

Wow. That is amazing!

If this takes off, that dude is gonna be RICH!

Well, yeah, there's that too........

 

[urbaindk]

The image gallery is amazing:
http://graphics.stanford.edu/papers/lfcamera/refocus/

Check out the water splash. Puts Edgerton's milk drop photos to shame.

 

[pixelninja]

Check out the water splash. Puts Edgerton's milk drop photos to shame.

While Edgerton's milk drop isn't that incredible by today's standards, the studies he did with motion capture were ground breaking. You also need to understand that he started his research back in the 20's. Can't really compare the 2 guys, except to say that they're both a lot smarter than me.

 

[LordOpie]

I see where he is going and it is very interesting . I just don't agree that it is a 4 Dimensional Field he is capturing. It sounds to me like he is capturing the entire 3 D world and then you can get to any 2d plane to have the picture you want . quite an interesting concept.

my thoughts exactly.

still, it's a great way to capture multiple layers of depth in a single instance of time. I bet sports photographers would love this camera!

 

[urbaindk]

Perhaps I should say that it "puts Edgerton's milk drop photos into historical perspective" or "takes Edgerton's work to the next level" then? Because I think it does.

All these cool sequence MTB and skate photos we see almost everyday are probably a direct result of Edgerton's influence wouldn't you say?

 

[stosh]

Looks like the 4D of light is just the blur tool in photoshop to me.

 

[LordOpie]

wouldn't 4d be this in motion?

Something like using this lens on a video camera and merging it with the technology that renders those 3d 'photos' in crystal that you see in mall kiosks?

that'd be cool... a 36" crystal in your living room capturing/displaying this 3d video.

 

[pixelninja]

Looks like the 4D of light is just the blur tool in photoshop to me.

If this first photo is the original:

then show me the tool in Photoshop that would allow you to create this photo:

Blurring images is easy. Un-blurring something so you get a true sharp image only happens on CSI.

 

[binary visions]

Interesting. Wonder if a similar effect could be achieved if you had a nearly infinite depth-of-field and some sophisticated software that would allow it to defocus all but a "plane" in the image.

Would require not only sophisticated software but probably some special image information passed from the camera to tell the software at what distance each point of light was from the camera, since it'd be hard for software to actually judge where the particular pixel is in real space.

Hmm...

 

[stosh]

If this first photo is the original:

then show me the tool in Photoshop that would allow you to create this photo:

Blurring images is easy. Un-blurring something so you get a true sharp image only happens on CSI.

Ha, I hate when they do that kind of crap on TV shows.


I just meant he had one picture with a large field of focus and from that he made different blury pictures. I was kidding man it's cool freaking stuff!!

 

[stosh]

If this first photo is the original:

then show me the tool in Photoshop that would allow you to create this photo:

Blurring images is easy. Un-blurring something so you get a true sharp image only happens on CSI.

Ha, I hate when they do that kind of crap on TV shows.


I just meant he had one picture with a large field of focus and from that he made different blury pictures. I was kidding man it's cool freaking stuff!!

 

[stosh]

That camera aint schit compared to a really really deep pool.

 

[N8 v2.0]

Dude!!!!

 

[ito]

Strangely enough after looking at his website I feel very stupid and doubt I will ever lead such an intelligent life or one that is so beneficial to humanity.....damn.

That is really freakin' cool though. Anyone look at the light photon thingy for making animation look like actual water? Computer geeks these days...it's amazing!

The Ito

 

[ZoRo]

Strangely enough after looking at his website I feel very stupid and doubt I will ever lead such an intelligent life or one that is so beneficial to humanity.....damn.

I feel the same way. The projects he is exploring for his PhD are amazing and greatly interesting.

 

[ZoRo]

Strangely enough after looking at his website I feel very stupid and doubt I will ever lead such an intelligent life or one that is so beneficial to humanity.....damn.

I feel the same way. The projects he is exploring for his PhD are amazing and greatly interesting.

 

[maxyedor]

Atualy if you just make the CMOS sensor in your digital camera more sensitive to light, you can shoot evggerything on f32 to get that expansive depth of feild; then just blur it in Photoshop to look the way you want, focused in the spot you like and there you go, same thing. You can accomplish this to some extent by using the exposure compensation already on your camera. Still a cool gadget, if it can be made cheap enough and get the same or better resolution as my EOS 1ds.

 

[Transcend]

Atualy if you just make the CMOS sensor in your digital camera more sensitive to light, you can shoot evggerything on f32 to get that expansive depth of feild; then just blur it in Photoshop to look the way you want, focused in the spot you like and there you go, same thing. You can accomplish this to some extent by using the exposure compensation already on your camera. Still a cool gadget, if it can be made cheap enough and get the same or better resolution as my EOS 1ds.

No, it can't.

You find me a tool in photoshop that will focus on a plane 7ft back, and then create DOF in every other direction...not going to happen.

 

[LordOpie]

I think he's suggesting that you capture the entire image in focus and blur out areas to create that effect.

But can you capture multiple layers of depth in focus at the same time?

 

[Transcend]

I think he's suggesting that you capture the entire image in focus and blur out areas to create that effect.

But can you capture multiple layers of depth in focus at the same time?

Right, but you can't simply do that. You could shoot with a pinhole camera and get infinite depth of field and focus from about 1" on out.

The problem is , there is no way to select a perfect plane to have in focus, and then add the proper DOF in all other directions. It is simply not possible. You would have to hand select crap in PS. It would look horrible, and take forever. This camera allows this. It is like the holy grail.

 

[LordOpie]

The problem is , there is no way to select a perfect plane to have in focus, and then add the proper DOF in all other directions. It is simply not possible. You would have to hand select crap in PS. It would look horrible, and take forever. This camera allows this. It is like the holy grail.

I can do it manually in PS... been doing stuff like that for 15 years, especially for jewelry catalogs... those people are nuts.

Slice o' life bro.

 

[Transcend]

I can do it manually in PS... been doing stuff like that for 15 years, especially for jewelry catalogs... those people are nuts.

Slice o' life bro.

There is no possible way you can select an exact focal plane. There was actually a big thing on this in the last photoshop professionals association journal thing I read at the library.

There is no way for PS to know what is at what depth in the photo. Sure you can add fake DOF fairly easily to say, a product shot. You can't do it on something as complex as a forest scene, or a rider on single track in the woods.

 

[SK6]

This whole thing makes me want to just sit back and scratch my nuts. In focus of course!

 

[maxyedor]

You can't select an exact focal plane, but you can get a very similar result by shooting in focus and then bluring portions correctly. That's not to say that this wouldn't be better, but you could just compose your shots, focused how you want them to be and negate this whole situation. 99% a shot is only going to look good with a specific amount of depth of feild, so when you change it the phot doesn't look nearly as good, thus why spend the exorbinate amount that this thing is going to cost?

 

[SK6]

The truth of the matter is, the end result can and only will be 2d. The issue is in regard to the selection of a plane and making it the focal point. That in and of itself is impressive as hell, because of the algorithms involved are just awesome

 

[binary visions]

This whole premise that you can duplicate a focal plane in photoshop is silly. If you spent many hours on one image you might be able to come close to it, but it would be an incredibly tedious, painful process - you're not just creating a blurred, unblurred image, you've got a stepped effect that slowly renders it further and further out of focus as it gets further away from the plane of focus.

For a product shot, okay, so it's easy - you just blur the background, since there is one definite subject and one definite background plane, and they are both parallel to the plane of focus. But that's not the way most shots are taken - in most shots, even if you have a definite subject and background, you have something in the picture that's perpendicular to the plane of focus, thus making it difficult to create your DoF effect.

Oh, and maxy, you forget that the more sensitive your film or CMOS or CCD or whatever you're exposing on is to light, the more noise you'll get. Simply making it the equivalent of an ISO3200 camera all the time just so you can get a deep depth of field is going to subject all of your images to some pretty extreme noise.

I think the idea here is awesome.. My speculation above was just wondering if you could pass distance information with the image to an image editing program - if your depth of field was big enough, you could then effectively use software to emulate whatever depth of field you want. You'd have to use a very small sensor or a very sensitive one of course, to get proper DoF, which would have image quality tradeoffs.

 

[firetoole]

can you make photography any more idiot proof?

"whats an F-stop?"

 

[H8R]

My head is spinning thinking of the filter and layer tricks that you can do to these images in Pshop. Split the DOF into 100 slices and then add filters to specific layers, or shift the layer slightly and do cool 3D effects by coloring the layers and wearing those glasses.

 

[BigMike]

can you make photography any more idiot proof?

"whats an F-stop?"

Buy a point and shoot camera, thats pretty idiot proof. Press the button

(and if you really don't know what an f-stop is, its the ajustment of the aperature. There are two MAIN things you physically ajust on a camera, the aperature, and the shutter speed. The Aperature is like an iris inside the lens that lets more light in at a lower "f-stop," (2.8 for example) and lets less light in at a higher "f-stop" (32). The effects of the f-stop is your relative depth of field, or DOF)

 

[Transcend]

This whole premise that you can duplicate a focal plane in photoshop is silly. If you spent many hours on one image you might be able to come close to it, but it would be an incredibly tedious, painful process - you're not just creating a blurred, unblurred image, you've got a stepped effect that slowly renders it further and further out of focus as it gets further away from the plane of focus.

For a product shot, okay, so it's easy - you just blur the background, since there is one definite subject and one definite background plane, and they are both parallel to the plane of focus. But that's not the way most shots are taken - in most shots, even if you have a definite subject and background, you have something in the picture that's perpendicular to the plane of focus, thus making it difficult to create your DoF effect.

Oh, and maxy, you forget that the more sensitive your film or CMOS or CCD or whatever you're exposing on is to light, the more noise you'll get. Simply making it the equivalent of an ISO3200 camera all the time just so you can get a deep depth of field is going to subject all of your images to some pretty extreme noise.

I think the idea here is awesome.. My speculation above was just wondering if you could pass distance information with the image to an image editing program - if your depth of field was big enough, you could then effectively use software to emulate whatever depth of field you want. You'd have to use a very small sensor or a very sensitive one of course, to get proper DoF, which would have image quality tradeoffs.

Bingo, thank you for paying attention.

Also, h8r is onto something. imagine what could be done if you could selectively output diff focal distances of the EXACT same shot and then layer them. Dear lord....

 

[Toshi]

Atualy if you just make the CMOS sensor in your digital camera more sensitive to light, you can shoot evggerything on f32 to get that expansive depth of feild; then just blur it in Photoshop to look the way you want, focused in the spot you like and there you go, same thing. You can accomplish this to some extent by using the exposure compensation already on your camera. Still a cool gadget, if it can be made cheap enough and get the same or better resolution as my EOS 1ds.

f/32 isn't the solution for two reasons:

1) diffraction causes softness
2) it isn't enough for more than a sliver of DoF when shooting macros

splat, i believe the fourth dimension of the whole "4D" business is the focusing distance itself? i still don't quite get how this whole micro lens + level recording business can be used to reconstruct the image... must read the paper ifself, i suppose.

 

[LordOpie]

I can select a focal point even at 1833 feet away on a flower in the Grand Canyon.

 

[urbaindk]

As you take the picture, I wonder if you could simultaneously send out a pulse of infrared light from the camera. If you measured the time it takes for the pulse of IR light to leave the camera and return, you could theoretically get depth information in a similar fashion to sonar or radar. If you could then somehow correlate this to each pixel captured in the photo you would know the depth of each pixel, from which you could generate a three dimensional image or do DOF manipulations after the fact. The trick would be to get spatially resolved information from the initial light pulse.