Phil's chart is lines per picture height. The 20 means 2000 lines, and the lines are alternating black and white. 2000 lines means only 1000 line pairs.
Chart only goes to 2000, so it is not showing more than 2000. 2048 can theoretically show 2048 lines, which is more than 2000.
Not exactly correct; if it were, the theoretical limit of a Bayer sensor would be 1/2 the one-dimensional nominal resolution, whereas in fact it approaches 1/sqrt(2).
As Phil explained it to me below ... It takes the Foveon chip only one pixel to resolve the black (0,0,0) or white (255, 255, 255), but it will take a Bayer sensor 2x2 pixels to resolve the same RG/GB. keeping that in mind, a 2x2 Foveon pixel area can be BW/WB, whereas a 2x2 Bayer area can only be black or white. So with 2048 pixles, you can have up to 1024 black/white pairs, or 2048 lines with the Foveon chip.
--
Huh? We're talking frequency here. One inscribes lines that are 1.414 times as wide as the pixel pitch taking into account the lens system magnification.
--
Example 1 at 400% in PS
http://www.coastphotos.com/pixel/example1.jpg
and full size crop
http://www.coastphotos.com/pixel/example1fsc.jpg
Example 2 at 400% in PS
http://www.coastphotos.com/pixel/example2.jpg
and full size crop
http://www.coastphotos.com/pixel/example2fsc.jpg
These are straight out of the camera. Zero sharpening in camera,
blowups are as displayed in photoshop. The first one, in
particular, is impressive. There's more aliasing going on in the
second, but it still shows line detail less than 2 pixels wide,
albeit a bit roughly
- DL
I don't disagree that the foveon will perform better than the same "resolution" (meaning image x-y) bayer, only that it won't be 3x better, and probably not even 2x better AS FAR AS detail. The image may well appear 2x better (or more) because it will be cleaner, with less aliasing, and better color, and this will affect our perception of sharpness (resolution).
This is not the 100% difference you implied earlier. To get back to the point of my examples, your "thought" example of a single line
--
--
Yeah, color complicates it, as does (I think) contrast. You'll notice in my example that the almost-single-line detail worked a lot better in example1 where it was basically grey, and low contrast.What if it were a black line running over a red, green or blue
background?
White is too easy
My interpolation of the same thing, but using Mitchell. No sharpening or any processing done other than a resize to 312x213 then back to 440x300, just like you said you did. Of course I used your JPEG image, did the processing, then recompressed, so there may be a few minor losses there.Here's my simulation of the difference between X3 and Bayer. This
is ONLY a simulation, the source image is actually a downsampled
D60 image. These are purely guesses.. we'll see the real truth
when I get a review camera.
Simulated X3 sensor:
I think the one I did using Mitchell interpolation is better. Still softer obviously. Still, note the thin middle line in the upper left crane. It has disappeared entirely from your image but is still present in mine. I'm not sure what changing the interpolation algorithm really means in this context however. But obviously you chose one to use, so I figured I might as well choose a decent one.Simulated Bayer CFA sensor:
The second crop was created by downsampling the first shot by 71%
and then upsampling it back to the original size.. This is
approximately what we'd expect to lose from Bayer interpolation.
Also bear in mind that the camera may sharpen the second image, we
haven't done that.
Yep. Before Intel we had IBM Power vs. DEC Alpha, with exactly the same thing going on. The Power did a LOT in each cycle, while the Alpha chose a very different route and did less, but enabled great clock speed scaling. But at least in that market most people used benchmarks. Now in the consumer market it is exactly that -- MHz sells. It's kind of like selling cars based on their maximum rpm.
Rock on baby
