ywenz
Veteran
You can't argue with this. WOW. Can you say. "virtual Leica lux bokeh" ?
http://graphics.stanford.edu/papers/lfcamera/
http://graphics.stanford.edu/papers/lfcamera/
jlw
Rangefinder camera pedant
ywenz said:
Yes, you can.
What they're doing is putting an additional microlens array over a very-high-density pixel sensor (medium format size) so that each image pixel is subdivided into several sub-pixels, enabling them to record the distribution of light energy over the area of the overall image pixel.
Then, using equations similar to those already in use for lens design and astronomical photography, they can interpret this energy distribution to calculate what the pixel would have looked like if the lens had been focused differently. (In principle, the same type of calculation could be used to eliminate lens aberrations, allowing the use of a lower-quality lens to produce high-quality photos.)
Very, very clever, and scientifically brilliant. However, if you've ever looked into any of the articles available about this technology on astronomy sites, you'll have found that getting these computerized calculations to work involves plugging in some basic assumptions about what the image should look like. I found one site with an interesting comparison of the same astronomical image reprocessed with different assumptions -- the differences were fairly startling.
That's not a big problem in astronomy because astrophysics provides a pretty good theoretical basis for assuming what should be in the corrected image. It's going to be a somewhat bigger problem in terrestrial photography, with the much wider range of possible subjects encountered.
How that most likely would manifest itself would be that one image would look great, and the next might exhibit inexplicable artifacts, or eliminate "problem" information.
I predict that this will be a great technology someday in, for example, camera phones -- it'll increase the flexibility of the product drastically without requiring additional mechanical complexity.
But for photography with a regular camera, focusing on the point you want in focus is still going to be a better overall solution.
ywenz
Veteran
What an image "should" look like is all relative to what you want it to look like right? Who said that the image they simulated has to be identical to what an actual Summilux lens produces? I'd love the opportunity to experiment with my personal looks..
For many applications focusing on a point before image capturing is constricting when compared to this method.
For many applications focusing on a point before image capturing is constricting when compared to this method.
jlw
Rangefinder camera pedant
I'm not saying it isn't an incredibly cool idea. I'm just saying it's going to have practical "gotchas" that aren't likely to be brought out in a scientific paper.
I think a more useful application than just building this into a camera (as most discussions of it seem to have emphasized) might be more what you seem to be suggesting. Suppose you could get a "light-field" camera that would save the energy distribution data as a raw file, and then process it afterward on a computer. This would let us do the same types of spatial adjustments as we can now do with exposure adjustments using today's raw files.
Where it would really get fun, as you suggest, is not just correcting focus, but changing the "look." This is already going on in music, where you can get audio plug-ins letting you simulate different types of vintage guitar amps, for example.
How cool would it be to open up an image file and be able to say, "Let's see how this would look if I had shot it with an uncoated 50/1.5 Sonnar...? Hmmm, a little too flarey -- how about a 1954 Summicron...?" and so forth.
The potential caveat to all this is the underlying and unexamined assumption that postponing the decision-making process until later is always better. In this view, the goal of photography (or anything else) would be to make the actual capture process completely decision-free -- then you'd just sit in front of your computer at your leisure applying endless interpretations until you came up with the one you liked best.
That's an appealing scenario, but (especially in documentary photography) there's always been a competing school of thought that says the best, or at least most authentic, decisions are the ones made in the heat of the moment. It makes sense, too -- the philosophy is that the photographer on the scene has access to a huge array of data from his/her surroundings, plus the immediate emotional connection of actually being there, and that all this contributes to better decision-making about how the picture should look.
Of course, in our current computerized culture that's not a popular viewpoint. If you wanted to get HEAVILY philosophical, you could speculate that our current enthusiasm for post-processing everything is because it offers a means of centralizing authority over our visual environment. The ideal is to turn the photographer into simply "skilled labor," charged with simply operating the data-gathering apparatus; the big, important decisions about how the photo should look are deferred until they can be made later by a higher-level executive.
Extrapolating far enough, we get a scenario in which the whole world is blanketed with continuous-video light-field cameras, feeding a central computer complex in which certified expert "artists" examine the stream of data, then process and generate the individual images that will be most effective in eliciting the desired consumer behavior.
Again, it wouldn't be that much different from the music business today, in which experts armed with sample libraries and non-linear editors synthesize the exact riff, lick, melody or whatever that will maximize sales of the artistic product. And they do fabulous work. So why would I rather listen to expensively-trained live people using pieces of wood and metal to play whatever it was that came into Mozart's head several hundred years ago?
I think a more useful application than just building this into a camera (as most discussions of it seem to have emphasized) might be more what you seem to be suggesting. Suppose you could get a "light-field" camera that would save the energy distribution data as a raw file, and then process it afterward on a computer. This would let us do the same types of spatial adjustments as we can now do with exposure adjustments using today's raw files.
Where it would really get fun, as you suggest, is not just correcting focus, but changing the "look." This is already going on in music, where you can get audio plug-ins letting you simulate different types of vintage guitar amps, for example.
How cool would it be to open up an image file and be able to say, "Let's see how this would look if I had shot it with an uncoated 50/1.5 Sonnar...? Hmmm, a little too flarey -- how about a 1954 Summicron...?" and so forth.
The potential caveat to all this is the underlying and unexamined assumption that postponing the decision-making process until later is always better. In this view, the goal of photography (or anything else) would be to make the actual capture process completely decision-free -- then you'd just sit in front of your computer at your leisure applying endless interpretations until you came up with the one you liked best.
That's an appealing scenario, but (especially in documentary photography) there's always been a competing school of thought that says the best, or at least most authentic, decisions are the ones made in the heat of the moment. It makes sense, too -- the philosophy is that the photographer on the scene has access to a huge array of data from his/her surroundings, plus the immediate emotional connection of actually being there, and that all this contributes to better decision-making about how the picture should look.
Of course, in our current computerized culture that's not a popular viewpoint. If you wanted to get HEAVILY philosophical, you could speculate that our current enthusiasm for post-processing everything is because it offers a means of centralizing authority over our visual environment. The ideal is to turn the photographer into simply "skilled labor," charged with simply operating the data-gathering apparatus; the big, important decisions about how the photo should look are deferred until they can be made later by a higher-level executive.
Extrapolating far enough, we get a scenario in which the whole world is blanketed with continuous-video light-field cameras, feeding a central computer complex in which certified expert "artists" examine the stream of data, then process and generate the individual images that will be most effective in eliciting the desired consumer behavior.
Again, it wouldn't be that much different from the music business today, in which experts armed with sample libraries and non-linear editors synthesize the exact riff, lick, melody or whatever that will maximize sales of the artistic product. And they do fabulous work. So why would I rather listen to expensively-trained live people using pieces of wood and metal to play whatever it was that came into Mozart's head several hundred years ago?
ywenz
Veteran
your posts are very entertaining
jlw
Rangefinder camera pedant
Thanks. I suppose I ought to start a blog or something -- then I could make possibly as much as $5 or $10 per month by subjecting people to idiotic banner ads while they were being entertained by my rambling rants...
Andrew Sowerby
Well-known
ywenz said:
Dunno if that's snarkiness, but I genuinely enjoyed those the above posts. Well thought out, well written.
sf
Veteran
Digital technology is superior . . .well, it is more advanced in some ways, but consider the fine evolution of earlier technologies - and their modernizations.
In some ways, film is more high-tech than silicon. It is the most evolved of the two - perhaps not the newest - but the most developed and perfected.
These micro lens arrays are interesting. . . the physics are beyond me, but it is interesting. Not my thing, though.
Will make for some monster files if they pull it off. BIG raw files. gigabytes with digital backs and large CCD cameras.
In some ways, film is more high-tech than silicon. It is the most evolved of the two - perhaps not the newest - but the most developed and perfected.
These micro lens arrays are interesting. . . the physics are beyond me, but it is interesting. Not my thing, though.
Will make for some monster files if they pull it off. BIG raw files. gigabytes with digital backs and large CCD cameras.
Trius
Waiting on Maitani
Frightening. You could "prove" that I killed someone when I was only a bystander. Or not even there.jlw said: