Slashdot Mirror
New Camera Sensor Filter Allows Twice As Much Light
bugnuts writes "Nearly all modern DSLRs use a Bayer filter to determine colors, which filters red, two greens, and a blue for each block of 4 pixels. As a result of the filtering, the pixels don't receive all the light and the pixel values must be multiplied by predetermined values (which also multiplies the noise) to normalize the differences. Panasonic developed a novel method of 'filtering' which splits the light so the photons are not absorbed, but redirected to the appropriate pixel. As a result, about twice the light reaches the sensor and almost no light is lost. Instead of RGGB, each block of 4 pixels receives Cyan, White + Red, White + Blue, and Yellow, and the RGB values can be interpolated."
"We've developed a completely new analysis method, called Babinet-BPM. Compared with the usual FDTD method, the computation speed is 325 times higher, but it only consumes 1/16 of the memory. This is the result of a three-hour calculation by the FDTD method. We achieved the same result in just 36.9 seconds."
What I don't get is calling the FDTD (finite difference time domain) analysis as the "usual" method. It is the usual method in fluid mechanics. But in computational electromagnetics finite element methods have been in use for a long time, and they beat FDTD methods hollow. The basic problem in FDTD method is that, to get more accurate results you need a finer grids. But finer grids also force you to use finer time steps. Thus if you halve the grid spacing, the computational load goes up by a factor of 16. It is known as the tyranny of the CFL condition. The finite element method in frequency domain does not have this limitation and it scales as O(N^1.5) or so. (FDTD scales by O(N^4)). It is still a beast to solve, rank deficient matrix, low condition numbers, needs a full L-U decomposition, but still, FEM wins over FDTD because of the better scaling.
The technique mentioned here seems to be a variant of boundary integral method, usually used in open domains, and multiwavelength long solution domains. I wonder if FEM can crack this problem.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
Foveon has 3 photodiodes per pixel, and theoretically should have the most accurate colors and sharpness by avoiding moire and interpolation issues with bayer filters. In practice, though, a lot of light is lost by the time it reaches the 3rd photodiode.
There is indeed white light because not every pixel has a filter over it. Many pixels pass the light through a hole to the pixel, while a neighbor pixel funnels red light (e.g.) to it. Thus, you get white + 1/2 the neighbor's red. You also get half the neighbor's red on the other side, resulting in white + red for the three pixels in a line.
Cyan is part of the color spectrum as a "subtractive color". What remains under each neighbor pixel when you strip away the red, is the cyan.
From what I can tell, this will not get rid of the need for the anti-aliasing.
...we've switched from calculating rggb values based on attenuated rggb values sensed, to calculating rgb values from sensing cyan (usually a color of reflected light with red subtracted, white+blue ?, white+red ?, and yellow (again reflected white light minus the blue spectral light.)
I can see the resulting files having better print characteristics, if the detectors sense to the levels close to the characteristics of ink used for prints, but I don't think that's going to help at the display the photographer will be using to manipulate the images.
And of course neither variety of photo image capture is comparable to the qualities of light that our rods and cones respond to in our eyes.
You never know...
In other words, technological superiority doesn't always win in digital photography.
This is very true, although the Foveon was superior in resolution and lack of color moire only - it terms of higher ISO support it has not been as good as the top performers of the day.
But the Foveon chip does persist in cameras, currently Sigma (who bought Foveon) still selling a DSLR with the Foveon sensor, and now a range of really high quality compact cameras with a DSLR sized Foveon chip in it. (the Sigma DP-1M, DP-2M and DP-3M each with fixed prime lenses of different focal lengths)
I think though that we are entering a period where resolution has plateaued, that is most people do not need more resolution than cameras are delivering - so there is more room for alternative sensors to capture some of the market because they are delivering other benefits that people enjoy. Now that Sigma has carried Foveon forward into a newer age of sensors they are having better luck selling a high-resolution very sharp small compact that has as much detail as a Nikon D800 and no color moire...
Another interesting alternative sensor is Fuji with the X-Trans sensor - randomized RGB filters to eliminate color moire. The Panasonic approach seems like it might have some real gains in higher ISO support though.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
"From what I can tell, this will not get rid of the need for the anti-aliasing."
You ALWAYS need antialiasing when you discretize.
Simply use three sensors and a prism. The color separation camera has been around for along time and the color prints from it are just breath taking. Just use three really great sensors then we can have digital color that rivals film.
Check out the work of Harry Warnecke and you will see what I mean.
Hey KID! Yeah you, get the fuck off my lawn!
Ironically, the last paragraph at Gizmodo somewhat answers your question:
What's particularly neat about this new approach is that it can be used with any kind of sensor without modification; CMOS, CCD, or BSI. And the filters can be produced using the same materials and manufacturing processes in place today. Which means we'll probably be seeing this technology implemented on cameras sooner rather than later.
Stay sentient. Don't drink bad milk.
"You ALWAYS need antialiasing when you discretize."
That's my motto!
This space available.
Interesting comments from both, but I believe you both missed the point. The real question is, which one of these methods, FDTD or FEM-FD, will allow optimal reprocessing in the frequency domain that makes my dinner look prettier with an Instagram vintage filter?
So when you print to your eight colour inkjet, what file format is your image stored in that has eight colour channels? What software are you using that supports it?
Note that in CMYK, which is the most by far the most popular "four colour" system (and is the one all those "four colour" printers use), black is one of the colours. That makes up for a shortcoming in the colour inks (which is not shared by camera sensors or displays) in which you can't make a decent black by mixing the colours. I suspect the eight colour printer is doing something very similar - mixing colours to give you a better (they say anyway) representation of the three colour additive system that your computer, camera and monitor use.
Besides, the vast, vast majority of people don't colour calibrate their monitors OR printers. Unless you do that regularly all the extra colour channels in the world aren't going to help you.
Is that one of those colors only women can see? Like mauve?
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
"From what I can tell, this will not get rid of the need for the anti-aliasing."
You ALWAYS need antialiasing when you discretize.
I think the word you are looking for is "quantize"
When working with designs meant for screen printing, the original artwork was done in RGB, then a team would separate the color channels (in Photoshop), one channel per ink to be used. They could technically do CMYK directly, but it didn't look good for a wide variety of purposes -- you can imagine a flat-filled cartoon character would be pretty much impossible. It would look a bit like comic book halftoning, probably. The shop would use that when they wanted to print Thomas Kincaide-esque sweatshirts for grannies. They would also use additional channels for things that weren't colors, like adhesive (for foil, usually) and clear inks.
I don't imagine that having more than three or four color channels is a new thing, or difficult to deal with. I would imagine even the prosumer technology would allow you to choose between various rendering intents. Probably the color separation is handled at the driver or device level, but TIFF, PDF, and DCS 2.0 (??) should handle extra channels natively.
A few more details on screen printing for those who might care: The actual screen printing process was not computer-controlled as a rule. The smaller shops I worked at printed a transparency which was transferred onto the screen by a photographic process, but the large one had a computer-controlled airjet "printer" that would knock out the design. Usually they would do a few samples by hand, to work out what ink and screen combination to use (different mesh sizes and ink thicknesses produce slightly different effects), and adjust finer details like when you would "flash" the shirt. That is, hitting it with a very high powered xenon lamp for a few seconds to dry the ink, before applying a new layer. You could do some interesting painterly effects with wet-on-wet ink; you can also make a hell of a mess that way. Flashing also tends to affect the color somewhat, especially for temperature-sensitive inks. After you get a few good samples, you send them off to the client as a proof. Then you would set up your automatic press for a run of a couple hundred. Color balance was something that the press operator kept an eye on after that point. After printing, the shirts are sent through a 400 degree open oven on a conveyor belt, for perhaps 10-20 seconds, to cure the ink.
Very fun job, the ink is messy as hell. I would still be doing it, but working with computers pays better.
Those who advocate genocide deserve every protection afforded by law, and none afforded by common human decency.
I gave an example of a printer with 7 colors+black, and yet you quibble over printer capabilities. Since reality is unrelated to your complaint, I'm curious what it is you are really complaining about. "There's no printer with 4 colours." "Here's one with 7+black" then you go off on some tangent about file support. I don't care. You are wrong. Everything you've said is wrong. I posted a link that proves you 100% wrong, and yet you keep insisting that reality is wrong and you are right. Yeah, I don't know where to go. If I answered your question about software, I'm sure you'd make up some other crap about paper quality or something.
Learn to love Alaska
Discretising is just quantising in the spacial domain!
Magenta is a combination of colours just like white isn't "in the colour spectrum".
Indigo/violet however is in the spectrum but as it's outside of the range of values which can be created with red green and blue we approximate it using magenta which is a mixture of blue and red.
So when you print to your eight colour inkjet, what file format is your image stored in that has eight colour channels?
You don't seem to understand the purpose of the colours or how colour is managed in a workflow. A file stored in your computer will have a certain gamut, if not specified this gamut is sRGB. Your printer also has a certain gamut. This is a function of the ink, colours it can print and the paper printed on. Colour management will take care of ensuring what you see on your screen will be reproduced on the printer providing the printer is physically capable of printing the colours in the gamut.
This is a quite common problem for instance with a CMYK printer which is unable to print any of the primary colours shown as red green and blue on the monitor. The result is a printer that prints a subset of the available colours a screen can display, but at the same time can print outside the gamut of your monitor too.
You don't need a file that has 8 primary colours to take advantage of the really wide gamuts 8 colour printers can print, you just need maths on your side. The ProPhotoRGB colour space works around this by defining the primary for green and blue as imaginary negative values which don't exist in reality. As such using red, green and blue primaries you can create for instance a colour that *almost* represents a pure cyan.
This is something that many photographers who print images already do. I think even the latest Photoshop comes setup out of the box to import raw camera files using ProPhotoRGB as the working colour space.
Besides, the vast, vast majority of people don't colour calibrate their monitors OR printers. Unless you do that regularly all the extra colour channels in the world aren't going to help you.
You don't know photographers very well do you? The vast majority of amateur and all professional photographers I've ever met calibrate their screens. Printer calibration is often not needed as the vast majority of photographers I know outsource their printing to someone else, and that someone else will typically provide them with the colour profile of their printer's last calibration to ensure accurate results can be obtained. Pretty much every printing company will do this for you, even cheap mass production ones like Snapfish.
You don't seem to know what we're talking about. Let me quote the OP:
"I've been hoping for 4-sensor cameras for ages. People only have three color sensors, but what those colors are vary a bit from person to person, and capturing 4 colors stands a better chance of getting images that look good for everyone."
Yes, more inks in your printer help it reproduce the RGB values that you capture with your camera, save in your files, display on your screen, and send to the printer. Just like in the example I gave, the K channel in CMYK helps make up for deficiencies in the mixing properties of the C, M and Y that don't let you make a proper black by mixing. Extra ink won't do squat to match extra colour information from a theoretical extra colour sensor in the camera though, because everything in between is RGB.
Yes, actually, I know lots of photographers. I calibrate my screen, and I use a printer I chose specifically because they do a good job of frequent calibration. Most professional photographers do. But if you haven't noticed, with the availability of digital cameras a LOT of people took up photography. Hardware screen calibrators are still a niche item, nowhere near as popular as cameras. In particular, Panasonic doesn't make any still cameras that are likely to be used extensively by professionals, so it's likely that even fewer people who shoot Panasonic would calibrate their equipment.
I'm not complaining about anything. I'm replying to your erroneous assertion (you DID read the whole thread before replying, right?) that the existence of printers with eight inks somehow means they'll be able to reproduce data from a hypothetical four colour channel camera sensor.
I do like your fake quotes though. Please indicate where I said "there's no printer with 4 colours." What I DID say was "Too bad you're displaying them on a screen or printing them with a process that only uses three colours." If you bothered to understand what you're talking about, or even read my comments, you'd realize that the process is indeed three colour. Even if you imagine a four colour camera sensor, the file you store the data in is three colour channel, the software you use to edit it is three colour channel, the screen you show it on is three channel and the data you send to the printer driver is three channel. IF you could somehow send the four channel data to the printer you might be able to reproduce some extra colours (which the vast majority of humanity probably wouldn't be able to see anyway), but probably not very well since all those extra inks are formulated specifically to help reproduce RGB.
That's wrong too. For example, if your image consists of widely spaced point light sources, it isn't low-pass filtered, but you still don't need or want an anti-aliasing filter to reconstruct the position of the point light sources. Not only don't you need an anti-aliasing filter, the image will look better without it. That's the case in astrophotography.
Whether you need anti-aliasing filters depends on what kinds of pictures you take, what you know about the scene, and what you are trying to get out.
Nope, it's not. Quantization is the process of taking a continuous valued measurement and rounding, truncating or otherwise cramming it onto a discrete scale. For example, taking the value 5.382... and recording it as 5.
I COULD have said "sampling." Sampling is measuring a signal at several points. The measured values are on the same scale the original was - if you're sampling sound with a microphone, for example, the samples are on a continuous scale. We almost always then quantize the samples, putting them on a discrete scale that's suitable to store in a computer. But sampling/discretization and quantization are two separate things.
Discretization is a term used more in math and statistics, but I used it here because it specifically refers to going from a continuous representation to a discrete representation. Sampling can also be done on a signal that is already discrete. We usually call that "resampling." If you're sampling a discrete signal and your sampling rate is equal to or higher than the original you don't necessarily need to low pass filter.
Discretization: http://en.wikipedia.org/wiki/Discretization
Quantization: http://en.wikipedia.org/wiki/Quantization_(signal_processing)
You might want to open your eyes and look in the 490–520nm range on a representation of the visual range of the EM spectrum.
To nitpick, that's actually not cyan. Cyan is a combination of green and blue light. The wavelength you're describing stimulates the green and blue receptors in our eyes in a way that looks (to us) identical to cyan, but it's not the same thing. Sort of like how violet (in the sense of being around 400nm) light stimulates the red and blue receptors in our eyes, similar to (but distinct from) certain shades of purple.
This becomes important when discussing things like optical filters. A cyan filter passes green and blue light. In other words, it is a red-blocking filter. This is very different from a filter with a bandpass of 490-520nm, which would also block most green and blue light.