Peter Spader
Veteran Member
Since people seem to have a desire to “talk” while we wait for the next SDxx camera from Sigma, and the “will they or won’t they and when” threads are getting so repetitive even humor can no longer make them interesting, I though I would revive another “golden oldy” (the issue of image quality) that has surfaced as a subtext in some of these other threads, since it is a topic worth of discussion and debate. Incidentally, I am not a scientist (though in my youth I had three years training as a Ceramic Engineer, one of the really “hard” sciences!) and so what I say below may well be inaccurate in some ways, and I expect people like JL and others will be able to correct any errors I present.
I start this thread because at the present time the CFA (aka Bayer) sensor based cameras have finally gotten large enough that they do surpass the Foveon sensors used in the Sigma SD9 and SD10 in resolution (as measured using B&W image tests.) In these tests one can see more “details” (of certain types---apparently the resolution supplied by Bayer interpolation varies so there may be things lost even with larger CFA sensors) than you can with the Foveon sensors. This does not surprise me since even though the CFA approach uses interpolation, sooner or later the area interpolated would become small enough that better details (of a certain sort) would become visible, and if the sensor is large enough you will have better resolution than the uninterpolated Foveon sensor. We appear to be at that point with some CFA cameras.
The problem is that this type of “resolution” is only one measure of image quality. It is a quantifiable term, it can be measured and quantified. The term “quality” is not. It is a holistic judgment that can be explained to a degree in terms of quantifiable terms such as resolution, but not reduced entirely to it. ( Resolution is one element of a good quality image. If an images resolution is low enough, then I doubt people would say it has the same image quality as an image showing much better resolution no matter how great its other elements may be. But resolution is not the only element that goes into the judgment of image quality.)
So why do people still find the images produced by Foveon/Sigma cameras captivating. I believe it is the quality of verisimilitude, the appearance of being real. I like the subtle shifts in the shades and intensities of color and the fine gradations of light, I see in the Sigma camera images. I like the clear demarcation of edges. I had been trying to capture and convey the feel of sunlight filtered down through trees in a forest for years with my Nikon 35mm film camera, and my first Sigma SD9 finally let me do it in the way I had been trying to capture. When I posted my first forest and falls picture someone commented;: “I feel as though I am there.” I have an A1 print (thanks to Laurence and Dominic) of one of my SD10 Nay Aug falls pictures in my office and someone looked at it and said, “I can hear the water running.” These are not avid photographers, incidentally.
Now I do believe there may be reasons why the Foveon/Sigma cameras produce images with this quality which has captured the imaginations of some of us. I suspect the fact that Foveon sensors, using their vertical stack of pixels (see definition of pixel on Foveon web-site, please) can capture both chroma and luma at each photodetection (X,Y) site does allow them to capture the subtle shifts of color and light intensity in a way no CFA array can do, and this may be why Foveon images have a sense of verisimilitude that is not totally tied to resolution. Another factor is the “edge sharpness” which the VFA approach produces, especially since there is no need for an AA filter (and the small AA effect the microlens on the SD10 sensor produce does not change this substantially.) The CFA (Bayer) approach must use AA to avoid nasty artifacts, but the Foveon sensor do not. Finally, the fact that the Foveon sensors do not convert data beyond their resolving power into “mush,” but instead produces “false” elements (on the resolution charts 9 lines are presented as 5) also helps. Yes it is “false” data if you are a mathematician or mapmaker trying to accurately record the number of line, but if I am looking at a landscape, being shown that “there are lines out there” is accurate information, and may be one of the “clues” our brains use to come to the judgment “that is real.” Verisimilitude is the “appearance” of reality, not reality itself.
So the fact that the largest CFA sensors have better simple resolution than the current Foveon sensors, does not condemn the current Foveon sensors from the standpoint of over all “image quality. Nor does it rule out the possibility that CFA sensors may not produce images with “verisimilitude.” (3D effects produced by effect use of contrast and DOF are also “clues” to reality.) Indeed eventually the areas sampled may become small enough it will become the functional equivalent of a much smaller Foveon sensor.
I have never suggested the better CFA based cameras produce poor images. My only direct experience has been full size TIFF from my nephews Canon 20D. They are good images, though to me they are overly “smoothed” (which is precisely what my young nephew prefers). As a VIS amateur photographer allergic to chemicals I am delighted the state of the art (both CFA and Foveon) sensors are now good enough I and so many others can now enjoy the “digital darkroom.”
Just be careful when making qualitative judgments.
And Seasons Greetings to you all.
Now go out and shoot some pictures and enjoy what you get, regardless of the camera you use.
Pete
I start this thread because at the present time the CFA (aka Bayer) sensor based cameras have finally gotten large enough that they do surpass the Foveon sensors used in the Sigma SD9 and SD10 in resolution (as measured using B&W image tests.) In these tests one can see more “details” (of certain types---apparently the resolution supplied by Bayer interpolation varies so there may be things lost even with larger CFA sensors) than you can with the Foveon sensors. This does not surprise me since even though the CFA approach uses interpolation, sooner or later the area interpolated would become small enough that better details (of a certain sort) would become visible, and if the sensor is large enough you will have better resolution than the uninterpolated Foveon sensor. We appear to be at that point with some CFA cameras.
The problem is that this type of “resolution” is only one measure of image quality. It is a quantifiable term, it can be measured and quantified. The term “quality” is not. It is a holistic judgment that can be explained to a degree in terms of quantifiable terms such as resolution, but not reduced entirely to it. ( Resolution is one element of a good quality image. If an images resolution is low enough, then I doubt people would say it has the same image quality as an image showing much better resolution no matter how great its other elements may be. But resolution is not the only element that goes into the judgment of image quality.)
So why do people still find the images produced by Foveon/Sigma cameras captivating. I believe it is the quality of verisimilitude, the appearance of being real. I like the subtle shifts in the shades and intensities of color and the fine gradations of light, I see in the Sigma camera images. I like the clear demarcation of edges. I had been trying to capture and convey the feel of sunlight filtered down through trees in a forest for years with my Nikon 35mm film camera, and my first Sigma SD9 finally let me do it in the way I had been trying to capture. When I posted my first forest and falls picture someone commented;: “I feel as though I am there.” I have an A1 print (thanks to Laurence and Dominic) of one of my SD10 Nay Aug falls pictures in my office and someone looked at it and said, “I can hear the water running.” These are not avid photographers, incidentally.
Now I do believe there may be reasons why the Foveon/Sigma cameras produce images with this quality which has captured the imaginations of some of us. I suspect the fact that Foveon sensors, using their vertical stack of pixels (see definition of pixel on Foveon web-site, please) can capture both chroma and luma at each photodetection (X,Y) site does allow them to capture the subtle shifts of color and light intensity in a way no CFA array can do, and this may be why Foveon images have a sense of verisimilitude that is not totally tied to resolution. Another factor is the “edge sharpness” which the VFA approach produces, especially since there is no need for an AA filter (and the small AA effect the microlens on the SD10 sensor produce does not change this substantially.) The CFA (Bayer) approach must use AA to avoid nasty artifacts, but the Foveon sensor do not. Finally, the fact that the Foveon sensors do not convert data beyond their resolving power into “mush,” but instead produces “false” elements (on the resolution charts 9 lines are presented as 5) also helps. Yes it is “false” data if you are a mathematician or mapmaker trying to accurately record the number of line, but if I am looking at a landscape, being shown that “there are lines out there” is accurate information, and may be one of the “clues” our brains use to come to the judgment “that is real.” Verisimilitude is the “appearance” of reality, not reality itself.
So the fact that the largest CFA sensors have better simple resolution than the current Foveon sensors, does not condemn the current Foveon sensors from the standpoint of over all “image quality. Nor does it rule out the possibility that CFA sensors may not produce images with “verisimilitude.” (3D effects produced by effect use of contrast and DOF are also “clues” to reality.) Indeed eventually the areas sampled may become small enough it will become the functional equivalent of a much smaller Foveon sensor.
I have never suggested the better CFA based cameras produce poor images. My only direct experience has been full size TIFF from my nephews Canon 20D. They are good images, though to me they are overly “smoothed” (which is precisely what my young nephew prefers). As a VIS amateur photographer allergic to chemicals I am delighted the state of the art (both CFA and Foveon) sensors are now good enough I and so many others can now enjoy the “digital darkroom.”
Just be careful when making qualitative judgments.
And Seasons Greetings to you all.
Now go out and shoot some pictures and enjoy what you get, regardless of the camera you use.
Pete
. The software side is long overdue. Results can be fantastic but the software itself is not very good.