05-21-2011, 09:05 AM
[quote name='Brightcolours' timestamp='1305944913' post='8711']
That could only happen if the CFA actually would just let through one wavelength, which is doesn't. Each "colour" lets through a wider band of the spectrum, making your example more or less purely theoretical...
[/quote]
Brightcolours, you've got some good information, and I for example appreciate you reminding us of the aspects of moire suppression problems that are not solved by a full-color-at-each-photosite sensor such as the Foveon.
But now you appear to be reaching for negatives about obvious good arguments for the Foveon sensor. As if you want to protect a preconceived idea that the Foveon sensor can't have any resolution advantage over a Bayer color filter array sensor. You seem here to have trouble acknowledging individually good aspects of an overall counter-argument...a basic tenet of effective rhetoric, to support the impression that your first priority is the facts. Of course we can all agree further that the highest and best use of a public forum is a search for truth with our peers, not to win arguments.
I happen to be both a professional photographer (albeit not a great one) and software developer. And have written a freeware raw file processor (i.e. a piece of software, albeit not a brilliant one). With a journeyman, but not ultra simplistic, demosaicer algorithm of course. But one need not be a computer scientist to realize that, when half the Bayer pixels are covered by a green filter, a quarter of pixels by a red filter, and a quarter covered by a blue filter, that in real green areas of an image (think foliage) a 14 megapixel sensor performs pretty much like a 7 megapixel sensor, resolution-wise. And in this example, in real red (think indoor lighting tending to this) or blue areas (think open shade's tendency) the sensor resolution begins degrading on the way down to 3.5 megapixels, resolution-wise. Similarly, in real yellow or cyan ("minus blue" or "minus red") areas you're heading down to 10.5 megapixels, and as the areas get more magenta ("minus green") you're heading South down to 7 megapixels. And man the perceived noise of every type goes up in real red, blue, or magenta areas when you don't have access to much data from the sweetest (most important to human eyes for luminance), green pixels.
Yeah none of the Bayer filters are perfect, and most colors and light sources in the world are not monochromatic, but I assume that a demanding user like yourself would be interested in even a 25% change in performance. And it's sure easy to imagine a scene or subject with say 75% of the green light gone that you'd see in a picture of a white thing in daylight.
Now at a higher and non-scientific level, it's true that modern cameras have such high resolutions that a typical 4x6 or 8x10 inch enlargement probably has little need of more than half or even a quarter of its nominal pixels to produce a nice image. And yes, the green filters over the various sensor sites aren't absolutely perfectly only letting in green light, etc. And it's true that any individual photo is not likely to have huge areas of pure magenta, or pure red, or pure blue, that knocks out the greens. And it's true that in a given year, there is no guarantee that the Foveon pixels are lower noise than competitor's Bayer array pixels.
But it would be easier for a camel to pass through the eye of a needle than for Bayer filter arrays to deliver full resolution in strongly colored areas. (Indeed in some of the very finest photos, the area of most visual interest is often the area of strongest coloration or are in the weirdest of light.) And Bayer array noise performance degrades in any area lit by light that especially is red, blue or magenta in cast (see above).
Yes, a given Bayer sensor could be so much more sensitive and noise-free than a Foveon in every way, that even with 1.5 hands tied behind its back, that Bayer sensor could whup the tar out of a Foveon with the same nominal resolution. Nah, not in strongly magenta, red or blue image areas, or in strongly bluish, reddish or magenta-ish light.
For every quibble you could make about the non-green light that makes it past the green filters, etc, there are a ton more arguments to make that a (say Foveon) camera with merely 50% less noise in strongly color-cast lighting conditions, or when photographing strongly colored scenes, would work at noticeably higher ISO with the same quality as a Bayer array camera. Because the resolution (to the extent that resolution is not degraded by noise which might be bad in say a poor-quality Foveon sensor) of a Foveon sensor never changes with strongly colored lights or subjects, and a Bayer sensor's resolution always changes in such cases.
Nobody's saying a Foveon camera has exactly 3 times the resolution of the same-geometery Bayer array camera. And nobody's saying that you ever take pictures in conditions or of subjects where the marginal advantages of Foveon sensors are significant. But neither will I let pass without protest your repeated implication that a Foveon camera has no better potential for resolution and noise in scenes that are either colored, or lit with colored (for example far from 6500K) light. Which is a whole lot of scenes.
Within the scope of this reasoning is still room for a given Bayer sensor to often and everywhere perform in a way that is not noticeably worse than a Foveon sensor. But by that logic one would have to also observe that an 8 megapixel camera can often and everywhere provide all the resolution that many a well-heeled user might notice or need. Similarly if you were a not-interested-in-max-quality user, I would agree that you might never notice the unique benefits of a Foveon sensor compared to a Bayer. But you seem to be just the kind of user who'd appreciate and acknowledge the relative, all other things being equal, greater resolution consistency and thus more consistently-smoother-looking images from a Foveon sensor camera of a given pixel pitch.
That could only happen if the CFA actually would just let through one wavelength, which is doesn't. Each "colour" lets through a wider band of the spectrum, making your example more or less purely theoretical...
[/quote]
Brightcolours, you've got some good information, and I for example appreciate you reminding us of the aspects of moire suppression problems that are not solved by a full-color-at-each-photosite sensor such as the Foveon.
But now you appear to be reaching for negatives about obvious good arguments for the Foveon sensor. As if you want to protect a preconceived idea that the Foveon sensor can't have any resolution advantage over a Bayer color filter array sensor. You seem here to have trouble acknowledging individually good aspects of an overall counter-argument...a basic tenet of effective rhetoric, to support the impression that your first priority is the facts. Of course we can all agree further that the highest and best use of a public forum is a search for truth with our peers, not to win arguments.
I happen to be both a professional photographer (albeit not a great one) and software developer. And have written a freeware raw file processor (i.e. a piece of software, albeit not a brilliant one). With a journeyman, but not ultra simplistic, demosaicer algorithm of course. But one need not be a computer scientist to realize that, when half the Bayer pixels are covered by a green filter, a quarter of pixels by a red filter, and a quarter covered by a blue filter, that in real green areas of an image (think foliage) a 14 megapixel sensor performs pretty much like a 7 megapixel sensor, resolution-wise. And in this example, in real red (think indoor lighting tending to this) or blue areas (think open shade's tendency) the sensor resolution begins degrading on the way down to 3.5 megapixels, resolution-wise. Similarly, in real yellow or cyan ("minus blue" or "minus red") areas you're heading down to 10.5 megapixels, and as the areas get more magenta ("minus green") you're heading South down to 7 megapixels. And man the perceived noise of every type goes up in real red, blue, or magenta areas when you don't have access to much data from the sweetest (most important to human eyes for luminance), green pixels.
Yeah none of the Bayer filters are perfect, and most colors and light sources in the world are not monochromatic, but I assume that a demanding user like yourself would be interested in even a 25% change in performance. And it's sure easy to imagine a scene or subject with say 75% of the green light gone that you'd see in a picture of a white thing in daylight.
Now at a higher and non-scientific level, it's true that modern cameras have such high resolutions that a typical 4x6 or 8x10 inch enlargement probably has little need of more than half or even a quarter of its nominal pixels to produce a nice image. And yes, the green filters over the various sensor sites aren't absolutely perfectly only letting in green light, etc. And it's true that any individual photo is not likely to have huge areas of pure magenta, or pure red, or pure blue, that knocks out the greens. And it's true that in a given year, there is no guarantee that the Foveon pixels are lower noise than competitor's Bayer array pixels.
But it would be easier for a camel to pass through the eye of a needle than for Bayer filter arrays to deliver full resolution in strongly colored areas. (Indeed in some of the very finest photos, the area of most visual interest is often the area of strongest coloration or are in the weirdest of light.) And Bayer array noise performance degrades in any area lit by light that especially is red, blue or magenta in cast (see above).
Yes, a given Bayer sensor could be so much more sensitive and noise-free than a Foveon in every way, that even with 1.5 hands tied behind its back, that Bayer sensor could whup the tar out of a Foveon with the same nominal resolution. Nah, not in strongly magenta, red or blue image areas, or in strongly bluish, reddish or magenta-ish light.
For every quibble you could make about the non-green light that makes it past the green filters, etc, there are a ton more arguments to make that a (say Foveon) camera with merely 50% less noise in strongly color-cast lighting conditions, or when photographing strongly colored scenes, would work at noticeably higher ISO with the same quality as a Bayer array camera. Because the resolution (to the extent that resolution is not degraded by noise which might be bad in say a poor-quality Foveon sensor) of a Foveon sensor never changes with strongly colored lights or subjects, and a Bayer sensor's resolution always changes in such cases.
Nobody's saying a Foveon camera has exactly 3 times the resolution of the same-geometery Bayer array camera. And nobody's saying that you ever take pictures in conditions or of subjects where the marginal advantages of Foveon sensors are significant. But neither will I let pass without protest your repeated implication that a Foveon camera has no better potential for resolution and noise in scenes that are either colored, or lit with colored (for example far from 6500K) light. Which is a whole lot of scenes.
Within the scope of this reasoning is still room for a given Bayer sensor to often and everywhere perform in a way that is not noticeably worse than a Foveon sensor. But by that logic one would have to also observe that an 8 megapixel camera can often and everywhere provide all the resolution that many a well-heeled user might notice or need. Similarly if you were a not-interested-in-max-quality user, I would agree that you might never notice the unique benefits of a Foveon sensor compared to a Bayer. But you seem to be just the kind of user who'd appreciate and acknowledge the relative, all other things being equal, greater resolution consistency and thus more consistently-smoother-looking images from a Foveon sensor camera of a given pixel pitch.