Continuation of the sensor working discussion.

[sidneylopsides]

I say this is what I mean Clicky (http://www.ddisoftware.com/reviews/sd9-v-bayer/)

I say they do not work like monitor pixels in reverse, they layout is different.

[sunystory@hotmail.com]

Foveon X3, the next generation cmos sensor will be adopted in samsung phones soon. V540 was the first example of the X3 introduced from Anycall brand:)

[sidneylopsides]

Foveon do work in a similar way to the pixels of a monitor, and give sharper images. Each pixel location on the sensor has a red, green and blue sensor. This does mean that the image size is 1.3 of the total number of pixels though, similar to a 3CCD system :)
Also, Foveon X3 is more of a layout that can be applied to sensor technologies, like the Bayer of SupperCCD patterns. :)

That image you have shown is SPECIFIC for SuperCCD (Fuji) or Foveon (some less-known brands) sensors and talks about interpolation, not real pixels

That image works for CCD and SuperCCD, not for Foveon. I belive you have mixed up standard (Bayer) layout and Foveon.

Normaly, no interpolation talking, 1 pixel, 3 sub-pixels.

Normally, 1 pixel=1pixel


PS: That last image show well MicroLens above SubPixels and silicon with contacts.

That last image shows the microlenses over individual pixels, not subpixels.


PSS: If everything worked as you said, how 2.000.000 pixels was 1632x1224 images, how 3.200.000 pixels was 2056x1532, 1.300.000 was 1280x960 ? :)

I'm not too sure what you are asking.

If, as you say, a normal CCD has 3 sub pixels for each pixel, how come Kodak and Sony have simply replaced the colour array filter on CCDs to give CMYK/RGBC instead of RGB? Wouldn't they have needed to add a subpixel?
And what about Foveon X3, why is that so special if CCD already has RGB per pixel area?

[rmcrys]

So much was said on the other post...

Well, I don´t have pictures so I´ll try to explain:

- for example, K750i sensor 2 Megapixels
- 2 Megapixels = 2.000.000 pixels aprox => produces an image aprox 1632 x 1224 pixels

this means:
1632 pixels [][][][][][][][][][]... in horizontal
and []
1224 pixels []
in vertical []


each pixel [] is constituted by sub-pixels that are colored elements

[(R)(G)(B)] or [(R)(G)(B)(C)] or [(C)(Y)(A)(N)] or whatever. The arrangement of those colored elements is varied. What is a sub-pixel? A sub-pixel is a material that when light strikes it, it returns a value of electricity. As to obtain colored pictures we need at least 3 colors, they have a colored filter for each color above them.

In case of K750i it uses a sensor production technic called CMOS (complementary metal oxide semiconductor) and has an effective 1.997.568 pixels, each one of those with 3 SUB-PIXELS => (REG), (GREEN), (BLUE) - RGB ===> 5.992.704 SUB-PIXELS

K750i has 1.997.568 PIXELS and 5.992.704 SUB-PIXELS

and for all of you people who doesn´t know, sensors also have lens for each pixel. Each pixel has a microlen above it to focus light on it´s sub-pixels.

Pixel diagram:
--------------
________ (silicon)
(R)(G)(B) (subpixel)
|______| (microlens)
^^^^
||||||
||||||
||||||
light
||||||
|______| (camera / mobile lens)
^^^^
||||||
external light

This is the basis of the all digital cameras´ sensors.

========================================

Then there are several kinds of manufacturing solutions: CCD (charge-coupled device), CMOS, SuperCCD, Foveon X3, etc.

On CCD there aren´t transistors for each sub-pixel so when the light stimulates each sub-pixel they return an electric value only readable at the end of the sensor. On CMOS there are more wires and a transistor for each pixel. Result => CMOS need less power, can "read" pixel-by-pixel which gives better color fidelity BUT, as near each sub-pixel there is a transistor it need more light to be electric stimulated => more noise!

The SuperCCD has and advantage over the above. It has honeycomb pixel format which allows better interpolation and an area for high and low-light captures => shadows are better handled.

Foveon X3 has another advantage => instead of sub-pixels disposed side-by-side, it has each sub-pixel one-above-another. This saves space and can return some good results.


But in the end and the bet of brands for CCD and CMOS, the developments make these two types of sensors the best. The others are good but have low developers and, as such, CCD/CMOS already catched them and passed over.

Low Cameras => Low cost CMOS because it´s easy / simpler to produce
Mid Cameras => CCD because it´s the best price/performance
High Cameras => Ultra High cost CMOS, almost nothing similar with low cost CMOS. Canon, for example, has produced a CMOS sensor where the transistors are in the back of pixels and EACH pixel has a barrier to separate from the rest => Low Noise, No Electrical Problems.

Hope I have helped.

[sidneylopsides]

So much was said on the other post...

Well, I don´t have pictures so I´ll try to explain:

- for example, K750i sensor 2 Megapixels
- 2 Megapixels = 2.000.000 pixels aprox => produces an image aprox 1632 x 1224 pixels

this means:
1632 pixels [][][][][][][][][][]... in horizontal
and []
1224 pixels []
in vertical []


each pixel [] is constituted by sub-pixels that are colored elements

[(R)(G)(B)] or [(R)(G)(B)(C)] or [(C)(Y)(A)(N)] or whatever. The arrangement of those colored elements is varied. What is a sub-pixel? A sub-pixel is a material that when light strikes it, it returns a value of electricity. As to obtain colored pictures we need at least 3 colors, they have a colored filter for each color above them.

In case of K750i it uses a sensor production technic called CMOS (complementary metal oxide semiconductor) and has an effective 1.997.568 pixels, each one of those with 3 SUB-PIXELS => (REG), (GREEN), (BLUE) - RGB ===> 5.992.704 SUB-PIXELS

K750i has 1.997.568 PIXELS and 5.992.704 SUB-PIXELS



That's what I mean. From what I have learned there are no sub pixels. I can't find any refernce to them on the internet.

The confusion lies in the definition of the term "pixel." Typically, in all except Foveon technology a "pixel" and a "photosite" are interchangeable terms because the final file produced has essentially the same number of display pixels as the sensor had photosites.

That statment says to me that the K750i will have 1632x1224 photosites.

[rmcrys]

Well, what is your understanding of photosites? If 1 photosite correspond to 3/4 colors and it´s the same as a "pixel", if a photosite is a color then K750 would need to have 1/3 of those pixels (because each pixel of a JPEG image needs 3 values of RGB to make the color)

[sidneylopsides]

A photosite is a single area that is light sensitive.
Each photosite corresponds to an individual pixel in the final image and photosites are often referred to simply as pixels.

My main point at the moment is, if most cameras use a Bayer colour filter array, which they do, how does your theory work? The method you describe is almost Foveon, and that is different by having 3 sensors in one pixel location.

for the final image the three colour channels are interpolated to fill in the pixel gaps where the other two colours are.
This (http://www.nezumi.demon.co.uk/photo/bayer/bayer.htm) page shows the whole process, the image he starts with is even a direct RAW from CCD before Bayer interpolation.

edit: This (http://www.bobatkins.com/photography/digital/raw.html) explains the process too.

Bayer pattern sensors Conventional digital sensors have their pixels are arranged as shown on the right, in a pattern known as the Bayer mask (or filter, or mosaic, or CFA for Color Filter Array), with two green pixels for each red and blue pixel. This is appropriate because the eye is most sensitive to green.

edit again: Check This PDF (http://dsplab.ece.cornell.edu/papers/conference/tifest_00.pdf) for an in depth explaation.

[rmcrys]

The process I said is the basis for every sensor. The disposal, once again has different forms, Bayer, Foveon, etc.

Your data is correct, so is mine. We are just talking about a little different themes: you are talking about specific sensor forms, I´m talking about how it works.

The link that talks about green being more sensitive for the human eye is not accurate, there are "2" green filters in bayer because the green is more difficult to stimulate by light and more difficult do render correctly. That´s why violet flowers are captured A LOT different from several cameras.

PS: if you read the PDF you´ll see "Due to hardware limitations (as I told in my post), CCD arrays in digital cameras do not capture the full Red [...] and interpolation is then used"

PSS: also in PDF says "Bayer array contains more green [...] in order to provide higher spatial frequency in luminance" (also confirms what I said about green)

Nice stuff you are getting. The problem is that each source have some things correct. If we mix some stuff together we get it all.

I think we have said is more than enough for people to understand. We aren´t giving a a degree on this stuff.

[sidneylopsides]

Actually, the human eye is more sensitive to green and uses it more than the other colours. The colour response is something like 68% green, 30% red and 2% blue. There is no reason why a CCD sensor would have more difficulty with green than any other colour.
I still say there are no sub pixels ;)

[rmcrys]

2% for blue? That´s a little low.