Slashdot Mirror
Wavy Lenses Extend Depth of Field in Digital Imaging
genegeek writes "On Feb 25 CDM Optics was awarded a patent for a new digital imaging system utilizing "Wavefront Coding" that produces images with 10-fold the depth of field of conventional lenses. The image itself is blurred until processed. Image examples are here."
can the images possibly be slashdotted already?
Basically what this is saying is that if I go out and get a new whiz-bang camera with this funky new lens, I will be able to take a picture almost as good as the pictures I take with my 30 year old Cannon AE-1, and not have the leeway of doing photo processing tricks in the darkroom.
Personally, I will stick to analog photography.
You say you want a revolution....
That wouldn't take long to saturate the processor. If it were flat html with images, it would just max out the network.
I hope the heatsinks work!
I couldn't help but think back to the problem with the Hubble Space Telescope, wherein after the launch they discovered that the mirror had not been properly ground to specification.
"Provided by the management for your protection."
Ah yes, I know this system well. I did my master's research in extended depth-of-field optics and came across this research which pretty much blew away what I was working on.
Here's a bit of background: in photography or laser scanning (point-by-point photography, basically), you always have a trade-off between depth-of-field and aperture size (as any photographer knows). Bigger aperture means shallow depth-of-field. However, a smaller aperture means lots of wasted light (imagine closing the aperture in your camera), and this means longer exposure times, and more importantly more NOISE in your images. This is true for digital, film, or photodetector.
So the "holy grail" is to keep the aperture open but still have high depth-of-field. This system depends on changing the phase of the light, instead of the amplitude (which is what you do when you stop down a lens to a smaller aperture). That way, no light energy is blocked and wasted.
Since the phase is changed, the resulting image on the CCD or film is fuzzy and has to be "decoded". You can think of it as "encoding" the wavefront in a special way that preserves the depth of field, capturing the image, and then "decoding" it into a sharp picture. It is really amazing. I hope it shows up in consumer cameras someday, it could completely change consumer photography since most "snapshot photographers" don't care about depth of field or all that stuff. It will also be great for medical and industrial imaging.
My system was sort of a hybrid between shading the aperture (instead of a sudden stopping of light, it gradually goes to black at the edge) and phase changes. Lots of people have been working on this problem over the years, but these guys really stripped the problem down to the essence and came up with a highly optimized solution.
Wavefront Coding?
Sound like something you geeks probably enjoy doing most when at the beach during spring break!
Didn't they say that the Hubble was broken? Now do we have to send up another crew to replace the inferior 'flat' mirror??? Make up your minds.
You try blowing up your 11MP image to 30"x40". I can do that just fine with my Kodachrome 64 on my Nikon FG camera.
Just because the base image quality may not be better (for 8x10 and larger from a 35mm sized camera, digital is so much better, but I like analog for 3X5 snapshots) doesn't mean the tricks and effects are neccessarily better.
Photoshop is great software, but no matter how much I try, basic manipulation (on b&w images particularly), especially brightness/contrast adjustment and dodging/burning, always gives me much better results under an enlarger. Same for exposure effects; Photoshop's solarize filter is good, but there's just some intangible warmth and...analog-ness to a well-solarized paper print. Maybe it's just the random scatter and size of the grain of film against the gridded regularity of the digital images, or the slight variation in quality across the print (not imperfect, but not...digitally homogenous), but for purely aesthetic ends, I have to go with film and paper.
Facts do not cease to exist because they are ignored. - Aldous Huxley
And I'm always leary of adopting a new technology that is monopolized by a single provider.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
A story...on Slashdot...about a patent...that's legitimate?
Does anybody remember the deep focus cinematography of Gregg Toland? How were those shots done?
If you post it, they will read.
Well they've been ./ed into oblivion.
Thinking about this from first principles, however.
There is only so much information flowing through the lens. By making it "wavy" aren't they just spreading the information out over a larger volume. In that case, they must be giving up either some contrast or some spatial resolution. Mayhaps someone more acquainted with the product can speak to this?
Mother nature is a b**ch, she doesn't give you anthing for free.
I'm suprised, the USPTO actually managed to issue a patent for something new and innovative and unique, rather than for something thats been common practice for a few years.
If I have been able to see further than others, it is because I bought a pair of binoculars.
"On Match 18 CDM Optics was awarded a patent for a new webserver system utilizing "Efficient Closure" that slashdots websites in one-tenth-fold the time. The website itself is blurred and inaccessible. We would give examples to you, but unfortunately they are all inaccessible"
There are at least two experienced photographers (Rob Galbraith and Michael Reichmann) who feel that the 11-megapixel Canon EOS-1Ds delivers images with detail exceeding that of 35mm and approaching (in some cases besting) medium format film. They've published some very interesting comparisons:
s p? cid=7-4833-4853
a s/ 1ds/1ds-field.shtml
http://www.robgalbraith.com/bins/content_page.a
http://www.luminous-landscape.com/reviews/camer
This may just change someone's opinon on how digital compares to film. I know it made me rethink the "conventional wisdom" that many more pixels are needed to reproduce film detail.
Cheers,
Jeremy
Their patenting the use of a lens in conjunction with post processing. Although the application may be new, the pieces are older. It would be the equivelent of patenting a computer chip in a shoe that keeps track of the number of steps you take.
End of Line.
have been in common practice for years. Now it seems all you need for a "legitimate" patent is that the "masses" don't know that other people are already doing some of it. Its actually not that different than what the other "bad guys" are doing.
End of Line.
The real problem there is dynamic range. Photoshop still works in 8 bits per channel, which is clearly not enough for any sort of exposure / brightness / contrast control. You need at least 16 bits per channel, preferably 32 (in floating-point format). Photoshop can load 16-bpc images but 99% of its tools are disabled until you convert the image down to 8-bpc. In other words: the 16-bpc mode is there just for marketing.
There are some interesting HDR (high dynamic range) projects, such as HDRShop, and these formats are also used in several high-end 3D renderers, but I don't think they will become mainstream until Photoshop adopts them.
Unfortunately, Adobe insists on minor updates instead of doing what Photoshop (and Premiere, and several other of their products) needs, which is a complete rewrite.
High-end 3D renderers also have very good "film grain" simulation (film grain is not just random noise, it has very specific characteristics), and other tricks that can make CGI "feel" almost exactly like traditional analog media. But again, this is not something you'll find in Photoshop.
RMN
~~~
havent been able to get to the site because of you lot bringing it down but... is this related to the technology used in NASA's docking cameras? I remember reading that they developed a camera that worked exactly as the /. story described, in order to combat the problem of losing focus on the target spacecraft during docking manoeuvers. The report I read was in New Scientist, probably 3 years ago?
I'd go and find it but NS archives are subscription only. I really ought to get round to subscribing, I buy it often enough...
-Baz
... and I can't find out because the site is /.'ed :-( :-) but some of us nature photographers do.
is this: Can this technology be used to control (not just increase, but also decrease) depth of field at image processing time? More specifically, can I get selective focus *after* creating the image? In criticizing my own work, I ususally wish I had openned up for *less* depth of field. I realize that sports photographers don't have this problem
Although Colorado University may never forgive me, this address has links to the research papers as well as more images: http://www.colorado.edu/isl/
That's because Photoshop & most digital cameras only use RGB colorspace (24 bits) which is a crappy color space, and one that we're currently stuck with because of our display devices.
High Dynamic Range Images use a higher bit depth (12 bits per chanel?). Many of the Nikon cameras can save out these 12 bit/channel images, which, with the proper manipulation software (HDRShop, others) can be used for much finer and subtler manipulation.
So, (math skills permiting), I make that out as 4096 levels per channel, as opposed to the current 256/channel in a standard 24 bit image.
It's still an RGB system, but it's a much better RGB system.
The next step is to get manufacturers on board & start making HDRI Video Cards & Monitors.
What were you expecting?
I understand the Slashdot has some silly qualms about caching websites which are linked in stories, but the benefit of caching screenshots, test images, etc. outways whatever it is that keeps /. from caching sites altogether.
The /. effect that is. :)
Telescopes are focused near infinity, but they have a very narrow depth of field. My telescope has a F ratio of F10 and a 2000mm focal length (although I usually use a focal reducer to bring it down to F6.4 when using 35mm film). This translates into a rather small depth of field.
Some more info from
Boulderdailycamera
Boulder startup gets deal with major optics player
By Anthony Lane
For the Camera
A Boulder-based startup, which makes technology that greatly improves the clarity of images through a lens, is poised to grow after signing a deal with one of the world's premier lens and microelectronics makers.
CDM Optics is a private company with sales last year of about $1 million, according to R.C. "Merc" Mercure, CDM's chairman and chief executive.
Next year, sales are expected to double with CDM's new partnership with the optical engineering company Carl Zeiss, a renowned manufacturer of microscopes, lenses and other instruments.
"The world's oldest optical company has joined forces with the most modern," said Ed Dowski, vice-president of CDM Optics.
The moving parts and multiple lenses of microscopes and certain cameras are precisely engineered to control aberrations and to produce a sharp image where someone wants it -- on a piece of paper, a slide or a computer screen.
Over centuries, scientists have devised ways to make sharp images of ever-smaller and more distant objects, but could do little to overcome the unchanging rules governing light and the formation of a focused image.
"There were no revolutionary changes in optics for 200 years," said Dowski.
CDM Optics produces an unusual type of "lens." Added to a standard lens, it produces images that actually appear blurry.
In fact, "There doesn't seem to be any part of the image that is more focused than any other," said Mercure, who was the co-founder of Ball Brothers Research Corp., which became Ball Aerospace.
A uniformly unfocused image may seem an unlikely goal, but after being digitally processed, the result is an image that is entirely in focus.
Mercure holds a poster with four pictures of a pack of crayons. Two were produced with a standard digital camera and the other two with a digital camera equipped with CDM's Wavefront Coding technology.
In one of the images from the standard camera, only a few crayons in the middle of the pack are in focus. To bring more of the crayons into focus, the photographer would have to decrease the size of the hole through which light enters the camera.
In the resulting image, more crayons are in focus, but it appears grainy as a result of less light hitting the camera's digital detector.
The difference between the two pictures produced with CDM's technology is more dramatic. The first is hazy -- it is an unprocessed image that would not ordinarily be seen.
In the second picture, all of the crayons from front to back are in focus without the graininess from the standard camera.
Dowski said applications for the technology that allows lenses to produce such images are numerous.
"You can either make lenses cheaper, sharper or both," he said.
Sharper images may be beneficial for many types of optics. A microscope, for instance, may magnify an object to 100 times its actual size with only a sliver 1 micron thick in focus.
"We can give a microscope up to 15 microns of focus," Mercure said.
One area in which this improved depth of field might be useful is in vitro fertilization. Ordinarily, a doctor produces a great number of embryos and monitors them for several days before implanting several. The goal is cause a successful pregnancy while minimizing the number of multiple births.
The problem is that after about three days, embryos are difficult to monitor with an ordinary microscope. The embryologist must guess which embryos are most likely produce a successful pregnancy.
Using Wavefront Coding technology, Mercure said, embryologists should be able to monitor the embryos for four or five days, thus reducing the number of embryos that must be implanted to have the same chance of a successful pregnancy.
The same increase in depth of field
"He is no fool who gives what he cannot keep in order to gain what he cannot lose."
Some of them are these days (wow! talk about low yield wafers!)
I doubt its that bad, since a camera can deal with a sparkling of 'dead' sensors, while pretty much any defect will kill a CPU.
autopr0n is like, down and stuff.
looks like those subscribers got to see the 'mysterious future' and /.'d the story before it went live. so i dont think ~20min until it got /.'d is some kind of new record.. almost makes you want to buy a subscription doesnt it?
Digital will never match the colour of slide film. It can't, by definition. It may be more vivid, due to some post-processing tricks, but it will never be as real or as authentic. Slide film captures the colour exactly as it was, whereas digital rounds it to the nearest bit. Slide film colour is as faithful and rich as the real thing.
Like woodworking? Build your own picture frames.
blehh damn preview button next to the submit button.
In case you were /.'d, most of the images from the CDM Optics website are also available here:
more images of increased depth
Since the phase is changed, the resulting image on the CCD or film is fuzzy and has to be "decoded". You can think of it as "encoding" the wavefront in a special way that preserves the depth of field, capturing the image, and then "decoding" it into a sharp picture.
When I first saw the article it sounded like the post-processing that is done to improve the focus of images that were originally taken out-of-focus. You can extract a lot of features by convolving an image with the inverse of the defocussing transfer function.
But doing this has a downside: It also brings to a point focus, or nearly so, the light from patches of a certain range of shapes. They weren't originally points - but photographing them defocussed made the same shape blur as a point light source would have, so the post-processing turned them into points. You extract features that would have been unreadable (like a license plate number), but also "sprinkle glitter and pepper" over the image.
Your explanation gives me some hope that the phenomenon might not occur with this system, due to a judicial choice of transfer function. But I'll wait to see the results from, not just a contrived demo, but a bunch of real-world pictures, plus some that were generated by a competitor who might have done a study of the physics and contrived, say, a "camouflage background" pattern tuned to spray crud all over the processed image.
(Such a pattern might make for interesting camouflage background and/or clothing. It would certainly screw up photos taken with the system, and might also produce strange results even with ordinary cameras. And any bets on whether human and animal eyes and brains use a similar trick?)
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
It looked like there were spaces in the preview, but I double checked and they were not in the input text field, so I just thought it was some funky Mozilla text rendering (which it's been known to do). As for making them active... an oversight on my part. Thanks for taking the trouble.
Cheers,
Jeremy
Wide angle? Hell, with a tripod to can make the field as wide as you want. Just rotate the camera while snapping shots. Paste together in photoshop.
Democrats or Republicans. They are both taking us to the same place and they are not afraid of us anymore.
whereas digital rounds it to the nearest bit.
Not exactly the brightest bulb in the flash rig, are you?
Actually, that's not true. Film captures color as realistically as the photochemicals can react to the incoming photons.
You can tell a great deal about the character of a man by observing those who hate him.
Maybe just the same info because I have not been able to get through to the original links.e ss/tech/27 bcdm.html
h tm
o loredf.ht ml
Here is a news paper article.
http://www.boulderdailycamera.com/busin
and another.
http://www.alteich.com/tidbits/t012802.
and some images.
http://www.colorado.edu/isl/intimages/3c
I think i saw something on this technology on Tomorrow's World ages ago, anyone else remember? I would check but the sites down..
This comment does not represent the views or opinions of the user.
Slide film captures the colour exactly as it was, whereas digital rounds it to the nearest bit.
This is what we refer to as "argument by bizarre definition".
Slide film captures color via photochemicals that change in response to light. Digital cameras capture color via sensors that signal in response to light. Saying that one is better "by definition" is patently absurd.
If slide film is inherently perfect, why are there so many different slide films with different color responses? If slide film captures color "exactly as it was", why is Fuji Velvia widely known for producing great landscape shots but murdering skin-tones? Slide film has all the same color concerns that any other capture method has -- good red response but poor greens, or great blues but muddy purples, for instance. Nothing is perfect, especially when the only real way to judge them is using the also-imperfect human eye.
I'm not basing my "better color" assertion on a bizarre definition of the abstract ideal. It's just my opinion, but I hold that my professional digital SLR, with little or no post-processing, produces better color than anything the film world has to offer. "Good color" is a subjective thing, and while you may disagree with me about that (cite examples please!), I stand by my statement.
ZFS: because love is never having to say fsck
Slide film captures the colour exactly as it was, whereas digital rounds it to the nearest bit.
Oh, dear.
Strangly, this too sounds like the guys who say that records sound "perfect" while CD's are limited by their bit rate. They're forgetting all the error inherent in the mass of the tonearm, distortion of the vinyl, etc, etc, etc.
The noise in analog systems is harder to quantify than the resolution based error of a digital system. That doesn't mean it's not there.
It may be more vivid, due to some post-processing tricks, but it will never be as real or as authentic. Slide film captures the colour exactly as it was, whereas digital rounds it to the nearest bit. Slide film colour is as faithful and rich as the real thing.
This sounds just like the whole 'Analog sound is warmer' argument I hear from some guy that just spent $15k on a stereo.
"We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
Douglas Adam's wrote that, it's nifty that you think that it's a great quote. So do I. You did not create it though.
This is not true. I noticed on your geeky art site that you use Fuji Velvia. It's a great film, but part of what makes its prints so beautiful is in the non-linear way in which it reproduces colours.
Chris Kuivenhoven is a thief, beware
I tend to like this site for photography help
http://www.usefilm.com
You can't get anything _after_ creating the image. You can digitally manipulate things by ephasizing or deemphasizing what's already there (sharpening, lightening, etc.), but there's no way to add data that just isn't there.
- Where film is tough to beat is in its dynamic range, not its spatial resolution.
True, that. It'll be interesting to see just how much Fuji's new SuperCCD SR does towards increasing the dynamic range of digital photos. Or what other advances might be made on the dynamic range front...Cheers,
Jeremy
>Slide film captures the colour exactly as it was, whereas digital rounds it to the nearest bit
Doh! Are you perhaps forgetting all that quantum mechanics stuff that's only been around for about the last 80 years?
In the end, you're just counting photon collision events - light is quantized. If you can get your CCD (and read-out systems) sensitive to that level (eg. see here), then nothing can be more authentic.
By definition!
I am talking a bit of trash there.
Unavoidable diffraction effects are certainly the "universal speed limit" when it comes to photographic resolution.
However, the Canon 1Ds of course could benefit from more pixels.
For one thing, it's got a Bayer? sensor, so it doesnt have Red, Green, and Blue sensing at every single pixel. That could be improved (like Foveon is trying), or the Bayer arranged sensor could be made more dense.
For a second thing, as with any Nyquist arrangement, you would want your pixels to be under 1/2 the size of the airy disc in order to fully resolve the lens capabilities on a sensor.
The biggest advantage to digital is by far the fact that it costs $0 per shot. (Well, maybe a few fractions of a penny for the electricity in the battery...)
So you can dynamically adjust your exposure settings as you shoot. It looks underexposed? Force an overexposure. No more need for an ultra-expensive spot meter.
In an optimal situation, low-speed film will blow away the best digitals. But digital shines in situations with unforgiving lighting, because you can see your results instantly and adjust. And once you get to the ISO 400 region, film quickly starts losing its resolution advantage. While it will take a 11+ MP camera to beat film rated in the ISO 50-100 range, it takes far less to beat ISO 400 film. 800 film? In many cases even a 3 MP digital will begin to win here.
retrorocket.o not found, launch anyway?
I was comparing to 35mm here.
Medium format, OTOH, isn't going away for a long time. Large format will live even longer.
retrorocket.o not found, launch anyway?
If anything, as other people posted, digital is closer to the "real thing".
One person mentioned that Fuji Velvia is great for landscapes but murders skin tones. This is because the sensitivity curve of a digital can be easily optimized, while it's very difficult to tweak the sensitivity and linearity of films based on chemical reactions.
As to rounding to the nearest bit - There's a lower limit in both electronic and film recording of the precision that a light level can be recorded which is distinguishable from noise. This is called the "noise floor" - Use enough bits, and then all the bit roundoffs will be well below the noise floor of even film media. (Which does indeed have a noise floor, just as digitals do. The nice thing about digitals is that with improved electronics and sensors, the noise floor of the sensor is dropping while film is staying the same. One of the things "pro" digitals are known for is having far less noise than lower-end digitals, and those improvements are constantly moving down to the consumer level.)
And for those that WANT the nonlinearities/quirks of film - All a camera manufacturer has to do is model the nonlinearities of major film types and then they can easily be emulated, just like guitar amps that use modeling techniques to emulate older units.
retrorocket.o not found, launch anyway?
Otherwise we are left with an optical viewfinder along side the real lens, and that has disadvantages.
Are you trying to tell me that the color of Velvia is real and authentic?
This already exists to some degree - Modern 3D cards have internal pixel pipelines of 16+ bits/color. The latest (GeForceFX and I think also the Radeon) have moved to internal floating-point representation, doesn't get much more accurate than double-precision FP. Even single-precision is insane compared to 8 bits/color.
Also, at the output stage, not much more than 8 bits/color is needed. At input, more than 8 bits/color would be useful for correcting exposure errors.
But for interim processing, many more bits are needed. Cumulative roundoff errors are evil, evil, EVIL. This is why most people consider a 24-bit DSP to be an absolute minimum for processing 16-bit audio. 16-bit is below the noise floor for almost all audio equipment to begin with, but if the processing path stays at 16 bits the truncation errors get compounded to the point where you've lost a few bits of resolution by the time you get to the output.
retrorocket.o not found, launch anyway?
digital imaging is saturated with marketing
tactics that rival the shennanigans of the
notoriously greedy redmond company we've grown
to hate. analog is here to stay and there is
enough room for digital
This post is probably* more useful than the original links to CDM optics website. In fact, there is a link on there to some of their papers.
From some of their "interactive" pages, (namingly this page), it seems as if they are using the "waviness" (I am still unclear about this) to do some amount of tomography.
This is the same thing that goes on when you get put into one of those CAT scan things at the hospital (I believe CAT stands for computer assisted tomography). It takes a volume (3-D) and projects various 2-D images. From these 2D images, one can then reconstruct the 3-D volume if enough projections are taken. If this is the case however, the downside is losing a great amount of spatial frequencies in the "depth" direction. These projections may be taken with a diffraction grating like thing (sort of a wavey pattern). The images seen on the above page look similar to pre-processed computed tomography images that I have done in some of my classes.
Of course, I could be wrong here too. I didn't read the entire articles mentioned above.
*I say this not knowing what is on the CDM website (unavailable) and guessing because the university is bound to divulge more info than a companyUnfortunately one of the largest contributors to the reduced noise in "pro" digitals is the larger sensor size--a feature that so far isn't finding its way down to the consumer level.
Cheers,
Jeremy
over the long run yeah its cheaper. When I got my 1GB microdrive in fall of 01 it was about $300, and flash cards run around $100 or so for reasonable storage. I suppose thats about 10-30 rolls of film before its 'free'.
http://www.economist.com/science/tq/displayStory.c fm?story_id=1476751
Really, thats all. Thats 4096 levels and you can't tell the gradient anymore. In fact a printer I've worked on maps 255 to 4096 via a LUT to help smoothe out the image... pity it can't take a 4096 image.
My understanding is that the color space used by digital cameras (256 values of red x 256 value of green x 256 values of blue) is both smaller and has less resolution than the color space of analog film. The color space for prints is more similar to that of digital cameras. This is one of the reasons that people doing digital film production generally use things like a 10bit log format (cineon) and 16bit per channel formats for final images.
Personally, I switched to using a digital camera over a year ago and I'm not looking back. I am very excited about the new wavy lens technology. However, I do wish computer technology would get to dealing with the color space issue. In my experience, blacks and greens in particular suck. I understand that some of the higher end digital camera can output images with more than 8bits per channel, but I don't know of any consumer grade cameras that do this or of any standard software packages that really take advantage of the extra bits.
Unfortunately it's a bit of a chicken and egg problem since there's no point in changing the cameras since most display and print technology can't display the extra bits and there's no point in changing the display and print technology since no one is producing higher bit depth images.
I've done some photography work, and I'm friends with a well qualified photographer... oh yeah, and my sister's dog once bit Ansel Adams or something like that..
Anyway, I think he was wondering if you could decode the image to focus on a specific range. Say the new lens gave you a depth of field from 3m to 12m and you just wanted a much narrower DoF. Could you decode the image to focus only on the near or far reaches of the DoF? He's not wanting to create new data after the exposure, just selectively use what's already there.
This technologyh has nothing in particular to do with the Hubble space telescope, other than "Wow, they both use lenses!" Get a fucking clue?
I wasn't referring specifically to consumer-grade digital cameras, which is what you are basing your argument on (just as the people arguing in favor of film are talking about professional-quality film and processing, not generic-brand film processed at the local drugstore).
My digital camera (Nikon D1X) captures 12 bits of color per channel, which is far, far more than the human eye can distinguish. Adobe Photoshop works with images of up to 16 bits per channel, and I consider that a standard software package.
Ultimately, of course, you're generally going to be producing an 8-bit-per-channel image, because even that is more than the human eye can distinguish. The extra bits just give you more room to play with before you downsample it.
ZFS: because love is never having to say fsck
is this one of those 5 line haikus?
Of course, he can't fit a 400mm telephoto lens to his camera :)
Any sufficiently advanced technology is indistinguishable from a rigged demo
--Andy Finkel (J. Klass?)
Depth of field is an optical property related to the focal length of the lens and the aperture used. Focal lengths on digital cameras are very short because the image sensors are smaller than a typical film frame. Even with large apertures, their depth of field is huge, making it next to impossible to achieve the shallow depth of field that gives you nice out-of-focus backgrounds in portraits, etc.
So basically, this is pretty useless for digital photography. Maybe it will be of limited use in fixed security cameras and stuff like that.
No, I didn't RTA. It's Slashdotted.
One could *maybe* do some SERIOUSLY heavy lifting to calculate distance info as one processed the image, but I'd be surprised if it was cost-effective. On the other hand, about a year ago, a video innovation was announced that would allow dynamic, realtime bluescreen, based on distance from the camera. JVC was involved, but I'm not finding the story... This worked by doing pixel-based realtime depth measurements.
If one were to take that additional information and store it, along with the pic, then you could easily do something like you want. In an advanced camera or at the desktop (photoshop or wherever) you'd simply declare: give me everything from 21 to 23 inches away, then run everything else thru a blur filter and put it in a layer behind my primary layer.
My first reaction in thinking about this was the market was too small to likely ever see this happen.
Then I started in on all the places that depth-based selective imagery would be interesting:
cleaning up touristy snapshots (getting rid of the tourbus in front of the Lincoln Memorial by cloning the stairway across the affected area), real estate photos, event/wedding photography, sports photography, ad photography... eliminating unwanted backdrop stuff (like power lines) would become a 'toss everything from 12m to 200m' tool. Heck, after a while, I couldn't think of an image I'd taken lately that I couldn't see some use for something like this.
Most of these are things that a digital photo editor does very well, though, so it may be unnecessary complications to the hardware. That especially goes for selectively blurring parts of your nature photos. I've got friends that do this professionally, and they regularly not only remove unwanted stuff but add stuff that was beyond the camera's field.
Last thought: it'd be cheaper and more expensive to do this depth-via-hardware compared to the video camera technology. Cheaper, because the need to do this 24/30/60 frames a second is gone. More expensive because you're no longer looking at the TV grid of 640x480 pixels, if you're preserving the depth info off a several-megapixel camera. That could make the distance sensor much more expensive.
Sounds to me like this lens would be great for high-resolution security surveillance camera. These applications typically use a tiny CCD so they can use ultra-short focal length. A 5mm lens has incredible depth of field at f/2.8. Much better then a 20mm wide-angle in 35mm format.
Now, with this technology, you could get an f/1.0 3 to 5 mm lens with a deep depth of field. Or something close to that effect. Combine it with super sensitive CCD and a good image processor and you could film things in near darkness without relying on infrared and IR spot lights to flood the area.
And then, there's TV and Movies. Let's not forget them. And cheap consumer digital camcorder which must take picture in very low light, self-adjust color temperature and then try to track a moving object while closing do f/1.4...
Who the hell cares about consumer P&S cameras? They already got a goddamn flash.
Alex
Crap.
/. setting 'lag time' so that you can actually follow the links of the stories when you read them a day later than the crowds.
Perhaps there should be a
--- Hindsight is 20/20, but walking backwards is not the answer.
Find me any analog photographer who shoots 1500 frames a year and hasn't come up with at least one major optimization in the film cost cycle. Your estimate on film costs assume that you're paying retail prices for film and processing, which is a little silly, isn't it? In the last year, I've shot over 200 rolls of film, each 36 frames. I spend about $2.50/roll on the film, and can process 4 rolls for something like $.32. I only print the real winners, and those only cost the price of paper (the cost of chems is negligible, ultimately, as demonstrated with the negative processing prices). So for 1500 frames, which is only about 40 rolls of film, I'm spending a grand total of about $100. At that rate, it'll take you a lot longer to make up the cost of your $1000 camera, by which point it will have been obsolete for several years.
Your call.
Your examples happen to be one of the draws of analog photography, btw. Film-based photographers love to think about which films get them the best colors for a given situation. Shooting on a cloudy day in Ireland? Try Agfachrome -- great greens and greys! Shooting portraits? I'll recommend Kodak for their excellent rendition of skin tones! Anything with sky or water in it? Fujichrome will take those colors away with their exagerated saturation curves!
You see, that's part of the art of it. Careful matching of film to situation, processing to exposure, exposure to visualized image, and then ultimately matching the finished print to what you visualized -- all of these things are critical parts of what make photography fun for many of us, and all of that goes away when you shift to digital.
I won't say you're wrong, but I will say you're looking at comparing something different.
Hello,
> My digital camera (Nikon D1X) captures 12 bits
> of color per channel, which is far, far more
> than the human eye can distinguish.
First of all unless your camera has active cooling I doubt there is a lot of information in the last 2 bits - Your camera may be capturing 12 bits but the usables one would be the first 9 or 10. Most camera's 8th bit is not that useful either. The rest is lost in noise of various origins.
Second of all the human eye can see more than 12 linear bits per channel. This is easy to prove. Right now for example I'm typing at a bright CRT screen in a darkened office while I look at some sunny outdoors. My eye can see all at once from decyphering the hand-written note on the post-it under the shadow of CRT to the brightest leaves outside, but not even film can capture that. I'm not even sure 16 bits per channel would be sufficient.
Also the RGB model does not capture all the colours visible to the human eye, this is well documented (CIE), and film is no better in that regard.
What people usually mean by `the eye cannot distinguish more than 8 bit per channel' is that most people cannot make a difference between a 7-bit encoded image and a 8-bit image in printed form or on a CRT screen. But there is much more to human vision than that.
Now if illumination were encoded on a logarithmic scale it might be different, but this is not what CCDs or CMOS chips capture.
Here are some more sample images: http://www.colorado.edu/isl/intimages.html
far, far more than the human eye can distinguish
Not completely true. If you do any processing of the images, only having 8 bits per channel is quite limiting, and color banding/quantization is not uncommon.
Additionally, the fact that camera negative film captures more color data than the print paper (or print film in the motion picture world) can capture gives you more latitude (and more artistic possibilities) when printing. You can print the "normal" ranges if you're doing something like wedding photos, you can print the lower ranges if you're shooting outside at sunset, etc. This is part of the art of shooting on film, and for a lot of photographers, part of the fun.
Adobe Photoshop works with images of up to 16 bits per channel
Yes, but it does it very badly. How many of the filters work with 16-bit-per-channel images? Not many. <shameless plug> check out CinePaint sometime </shameless plug>
You probably mean BPC (per channel), not BPP (per pixel).
Anyway, it's not about telling the gradient, it's about preserving information. Even 10 bpc is more than enough for the output medium (even 8 bpc is more than we can distinguish, for some colours). The problem is the precision used for the calculations and the amount of information about the source. Traditional chemical film captures a lot more information than what you can see on photographs (as anyone who's into photography knows very well). That information is then either compressed into the dynamic range of the photograph or the ends are cut off (usually both). That's why you can make two (or a hundred) completely different photographs from the same negative.
Regarding your example, (mapping a scale of 255 values onto a scale of 4096), unless you also increase the resolution (and interpolate the new pixels), there's nothing to be gained; you are simply multiplying the values by 16). And besides, printers only use a handful of inks, when people say a printer "can do 1440 dpi", each one of those dots can only have one of four or five colours. To create the illusion of intermediate colours you need to consider groups of dots, not each individual dot.
RMN
~~~
Most photographers want LESS depth-of-field than the current crop of digital cameras provide.
Only amateurs want "everything from here to infinity" to be in-focus.
The advantages of selective depth-of-field cannot be understated. The ability to have the background be completely soft and have the subject be the only thing in sharp focus (thereby drawing the viewer's attention to it) is a huge advantage of film over digital.
For example, on Attack of the Clones, the guys at ILM actually had to process the images to give them less depth-of-field, because the cameras couldn't get as little depth-of-field as the cinematographer wanted.
*sigh* thats what I get for reading all of the responses, getting irritated, and typing out a short response because a long one wouldve been wasted. With 12 bits per channel, you are correct, you get 36 bits per pixel, encoded on a 48 bit per pixel setup. Wasting space...
As for the interprolation, well, the printers can create very smoothe contoured prints, but you need to 'notch' the levels via the 8 bit LUT. Thats what I meant- there is no additional data created (although frankly you could perform some gaussian distribution analysis to figure out how to 'blur' over the extended range).
Some eyes are better than others at detecting 8 bit problems. Just look at a person's eyes from your digital camera, if you get the chance, and look at the iris.
Really...? Did you even read what I wrote? Let me repeat it:
"Photoshop can load 16-bpc images but 99% of its tools are disabled until you convert the image down to 8-bpc."
Try enabling the 16-bpc mode. Then try to paint. Or apply filters. Or use layers. Understand now?
Pretty much the only thing you can do with 16-bpc images in Photoshop is look at them. But since your graphics card is probably using a 24-bit mode, you're really looking at an 8-bpc image anyway.
Besides, 16-bpc != HDR. Proper HDR processing needs floating-point channel support, otherwise you're still establishing hard limits for the dynamic range.
I've been using Photoshop (and several high-end programs and systems) professionally for several years. Photoshop is still ubiquitous but very few people nowadays can afford to use just Photoshop. There are some free / shareware programs that clearly surpass it in terms of quality and features (support for HDR / FP and parametric editing being the main issues). The two things keeping Photoshop as a "must-have" application are its ease of use and existing user base. I hope Adobe doesn't fall asleep at the wheel as happened with many other software (and hardware) companies.
RMN
~~~
Appears to save in an extended bit-range format:
JPEG (EXIF 2.2)
TIFF (8-bit)
RAW (12-bit)
From what I understand, this raw format is 12 bit/channel, with the extended luminance range that implies.
I'm actually looking at this camera to capture HDR images to use with Lightwave's HDRI renderer.
(of which there are some sample on the page you link). More study is needed.
What were you expecting?
Intel engineering seem to have misheard Intel marketing strategy. The
phrase was "Divide and conquer" not "Divide and cock up"
-- Alan Cox, iialan@www.linux.org.uk
- this post brought to you by the Automated Last Post Generator...