understandign actual pixels

[WmAx]

[Do you happen to know what Fuji means by S and R pixels?
Here is their description from their web page for the S3

S and R photosites are photosites with different native sensitivities. A SR sensor as one S sensor and R sensor at the same spatial position. Therefor, since it takes 6 million R + 6 million S pixels to equal a 12 MP SR sensor, their are only 6 million spatially unique positions. Therefor it will be effective as a 6MP camera as far as resolution is concerened. The S and R pixels work at different senstivites and take two unique pictures simultaneously. They work 2 stops apart from each other(4x llight difference). The two images are later combined/superimposed in a in order to capture a broader scene of light then a normal sensor would be capable in a single exposure.

I'm also understanding that pretty much all cameras interpolate except maybe Sigma's. The S3 also does this however because of their photosite layout they basically do a much better job of it right?

I can not speak about how 'good' interpolation may be performed by a specific camera. However, it should be noted that the interpolation is not the same as 'resampling', as you may be thinking. It is creating alternate values using surrounding(color) data that is available, which is basicly the definition, though. The SCCD based camera(like S2, S3) perform an additional form of interpolation, as I illustrated above, in addition to the normal color interpolation(demosiacing).

Here is a crude example of a bayer sensor:

R=red
G=green
B=blue

RGRG
GBGB
RGRG
GBGB

Each of the individual R, G, and B sites is used by the demosiacing algorythym to estimate/solve the approximate real color that each pixel should compose. The result is a R, G and B at each of those points.

X=Red, Green and Blue

XXXX
XXXX
XXXX
XXXX

Of course, this leads to color errors and reduced resolution, as compared to if you had a R, G and B sensor at every point originally. This is why a 3.5MP Foveon sensor resolves greater detail then a bayer sensor. It has R, G and B sensor at every spatial sampling point. No demosiacing is required.

Is the aliasing you talk about what causes the purple hue that is sometimes seen at areas of high brightness with darker backgounds such as a flashes reflection off something shiny such as chrome?

No. The purple hue is a result of other factors.

Here is an example of an aliasing byproduct:



Look at 14 mark, their are 9 distinct lines. Look at the 20 mark. Their are now 5 or so lines and they are thicker then the top lines. IN reality, the chart had 9 distinct lines to the 20 mark that gradually got thinner. When the spatial frequency of the chart exceeded the nyquest limit of the sensor(which has not anti alias filter), the sensor misread the multiple high frequency lines and falsely displays a lower frequency modulation. Result is a false pattern. This will happen on any pattern that exceeds the spatial frequency of this sensor. This is only an example that lets you clearly see one effect of aliasing. The other effect is jaggies(staircaising) on some high contrast edge transitions on curves and diagnols.

-Chris


[/img]

[JerryF]

This is wonderful information, but should all of this be in the "Newbie Help" section?

[Caveboy_Ken]

Perhaps not as it is very detailed however it is pixel related, I'm a learning newbie and am very grateful for everyone's input in helping me to understand. Thanks.