missionphoto
Offline
• •
Upload & Sell: Off
|
 p.2 #17 · The Foveon Concept in Visual Form | |
Pondria wrote:
Missionphoto, Thanks for the perspective !!!
I thought I was the first one that insisted SD10 is a 10MP camera. Yeah, I got some reactions, too, from both Foveon and Bayer camps 
OK, have we now established that SD10 is equivalent to 10MP Bayer Camera for naive MP-oriented discussions ?
Then my old question should be asked again. According to DPReview and Imatest, SD10 gives the resolution equivalent to 10D, which is a 6MP camera.
Unfortunately, dpreview's resolution tests are worthless for this comparision. Phil knows it. I showed him his error years ago. The "error" is actually an intentional skew from way back, get this, on or around 1969. Bryce Bayer is one of the few people to really understand the differences between Bayer mosaic sensors (his namesake) and color sensors (at that time, film). I'll get to that in a sec but first a quick aside...
The history of this is more interesting than the imagery. No one has assembled it in its entirety, to my knowledge, but it goes something like this... In the late 60's Bell labs invented the monochrome digital sensor. A key scientist, some say he invented it, was a guy named Carver Mead (you might recognize the name, he is the founder of Synaptics, inventor of touchpads, and later he founded an obscure company named Foveon. Inc.). Bryce Bayer, I believe of Kodak fame, was the first to patent a color mosiac pattern that could sit in front of that monochrome sensor to extract a color image.
As the Bayer mosiac sensor was develped, it became obvious that in order to get color from a monochrome device, you had to lose 67-75% of the B&W resolving power. Why? Simple. Using the common color model, you need three simultaneous color exposures (Red+Green+Blue) to get a color image. There are lots of ways to do that. The most obvious way is with separate exposures, like old fashioned RGB film plates, or a 3CCD or 3CMOS design where a prism divides the light into RGB and you have a dedicated sensor for each RGB image. So to understand everything you need to know about Bayer resolution, just remember this: a Bayer sensor is nothing more than a monochrome sensor which is subdivided into 3 smaller, simultaneous RGB exposures. It really is that simple. Knowing the pixel dimensions of each individual RGB exposure is the key to knowing the full color resolving power of the sensor. Full color resolution is the resolution of each individual color component. It really is just that simple.
The AA filter is oft discussed but theoretically it doesn't matter, in so far as its primary function (other than it adds another optical layer and that is never good, so do microlenses, dust protectors, filters, etc). The AA filter doesn't matter much because its influence is designed to be limited to only adjacent pixels. In other words, it doesn't blur anything within any single RGB color exposure, since none of the pixels in say, the red exposure, are adjacent to one another. I know that is an oversimplification/approximation, but you get the point.
So a Bayer sensor is a true 3CCD design, of course. One red, one blue, one green CCD, but each of the RGB CCDs is intermingled with the next. Thus it's full color resolution is no different than any 3CCD design with that number of pixels in each RGB exposure. How could it be otherwise?
Back to dpreview's error. Developers of early digital cameras knew that color sensing sub-divided their large sensor into 3 little ones, but they wanted to compete well against film. So they did something very clever and they have been fooling the Phil Askey's of the world ever since. They declined to test color resolution. They used a B&W chart to inflate their Bayer sensors' performance relative to, then, color film, and now by happenstance, Foveon.
The "trick" is as simple as this. What is the one color that all three mingled RGB exposures have in common? The answer is black. No-light can be detected equally well by all three exposures, because they all can register nothing. The B&W digital test pattern was born. It is the only test that tests the Bayer color sensor as if it is still a monochrome device.
Like lemmings trundling off into the ocean, pro reviewers never thought to ask themselves the glaring question, "Why are we testing color cameras with B&W images?"
That is why.
Does it mean that Bayer is superior with the same number of photo sensors ?
Nope, given the same number of sensors the two are, overall, the same. That is the true, shocking conclusion to this age old debate. There are some fairly significant differences within individual color channels, but that balances out exactly in terms of total resolution, because the total sensor count is the same.
So which is better overall? Neither. But each holds substantial advantages over the other depending on the colors in the scene and what you prefer in a photograph. The biggest diefference to me is that Bayers have imbalanced color channel resolution, so you lose some 3D effect since varied RGB color channel resolution/blur competes, optically improperly, with optical lens blur that tends to protray depth to the viewer.
Edited by missionphoto on Oct 01, 2006 at 10:44 PM GMT
|
