Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
06-06-2009, 06:01 AM   #16
Inactive Account




Join Date: Dec 2008
Location: Ames, Iowa, USA
Photos: Albums
Posts: 2,965
QuoteOriginally posted by Quension Quote
...
One issue is that sensitivity for the colorless pixel would need to be different from the filtered ones; you would have to nearly triple the dynamic range to allow for complete white without losing sensitivity to pure green. Increased dynamic range is a goal sensor manufacturers are still chasing, and sticking with RGGB (or other individual color filters) would allow improvements to be applied to the entire image instead of being limited to subtractive color detection in part of it. It might not be considered worth trying until sensors can exceed our eyes' dynamic range and there is excess to play with.
Good thoughts; let's look at some numbers for constant dynamic range sensor elements with different filters.

Say we've a flux of 3 thousand white (1000Red/s+1000Green/s+1000Blue/s) photons per area-second and a sensor (underlying a filter element) becomes saturated when it has received a thousand photons/area.

It would take 1 sec to saturate the RGGB sensor array; 4000 photons total would be collected during that time (1000 for each of R,G,G,B).

Now consider the RCCB sensor; the same 3 thousand photon per area-second flux would take 1/3 second to fill the clear cells, during which time the R & B cells would each have collected 333 photons; a total of 2666 photons in 1/3 sec.

The RCCB sensor's saturation based ISO would be 3X that of the RGGB sensor.

The ISO for pure white would be the same as that for pure green (the meter doesn't know a photon's color....pure green light would fill the clear cells in the same time that pure white would fill them.)

The luminance S/N ratio would be about 40% better for the RCCB sensor (comparing the sensors at the same exposure time.)

Whether the gain in ISO/noise is worth the loss in color information precision is a separate question; I think that may be the case - color information precision is much less important than luminance IIRC.

I guess the engineers who design these things have got a lot to think about.

Iowa Dave

EDIT: consider the ISO at constant luminance S/N ratio (based on time for same total photons collected):

RGGB ...4000 photons takes 1 sec
RCCB....4000 photons takes 1/2 sec

So the noise based ISO is two times faster for the RCCB filter array - this is probably the best comparison regarding image quality.


Last edited by newarts; 06-06-2009 at 02:27 PM. Reason: added noise based ISO comment
06-06-2009, 02:32 PM   #17
Senior Member
jslifoaw's Avatar

Join Date: Apr 2007
Location: Toronto/Victoria
Posts: 262
Would having a pixel-level "3CCD" style filter work? Or is there no sensitivity gain by doing so?

I.e. for a single "pixel", have three sub-pixels where some lens above it distributes the light into the R, G, and B buckets accordingly.

Almost like Fuji CCDs where at a single pixel you have two buckets with different reponse.
06-06-2009, 02:47 PM   #18
Site Supporter
LeoTaylor's Avatar

Join Date: Feb 2007
Location: Connecticut
Photos: Gallery
Posts: 679
QuoteOriginally posted by Quension Quote
... It might not be considered worth trying until sensors can exceed our eyes' dynamic range and there is excess to play with.
You bring up an interesting point which I do not know the answer to: What is the dynamic range of the human eye. I'll throw out a few numbers that I have read but have no way to verify the first item.

1. The human eye can see 64 gray levels (6 BITS). Don't be confused by the eye opening and closing the iris to increase the range visible. That is like opening and closing the aperture of a lens to move the sensors usable range. It is NOT a function of the sensor, whose job is to resolve brightness instantaneously without external help.

2. A typical RGB computer display has 256 Gray levels (8 BITS). I presume whomever came up with item 1 found the eye can't readily detect all the levels a computer could display?

3. My Pentax K110D has 12 BIT output which implies 4096 Gray levels though it may not actually achieve that potential.

4. A SBIG astro camera has 16 BIT output and can resolve 65,536 Grey levels.

I just used PSP-11 to create three squares with RGB levels of 128,128,128 129,129,129 and 130 130 130. I'm hard pressed to say I can see a difference and this is the easy mid range brightness. I tried 0,0,0, 1,1,1, and 2,2,2 and they all look black to me. My monitor is calibrated with the Spyder2 device.

So, is the eye really better than a CCD sensor?
06-06-2009, 02:53 PM   #19
Inactive Account




Join Date: Dec 2008
Location: Ames, Iowa, USA
Photos: Albums
Posts: 2,965
QuoteQuote:
Would having a pixel-level "3CCD" style filter work? Or is there no sensitivity gain by doing so?

I.e. for a single "pixel", have three sub-pixels where some lens above it distributes the light into the R, G, and B buckets accordingly.
Are you thinking of something like a little prism that spits the light into its component colors then directs the results to different detectors? - That'd be great but likely hard to do. In effect that's what the Foveon more or less does by differential absorption of the light's components - a great idea.

Iowa Dave

PS I'm sorry if I was sloppy in my description of the RCCB filter array......

Each pixel of a standard digital sensor is subdivided into 4 smaller pixels; one of these smaller pixels is covered by a red filter, two by green, and one by blue (called a Bayer Array). I am describing a system in which only two of these sub-pixels are filtered; one by red & one by blue.

06-06-2009, 05:55 PM   #20
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
QuoteOriginally posted by newarts Quote
It seems to me a color array of blue-white-white-red might be an improvement over the Bayer red-green-green-blue in use.

The RGGB array makes white by sum. A RWWB array would make green by difference (2G=W+W-3R-3B).

There is no intrinsic color advantage to this, but there would be a significant shot noise improvement.

Roughly speaking, an exposure of white light photons incident on a RGGB array gives 4 equal count outputs; one for red+ two for green+ one for blue. But for a RWWB array the same stream of photons gives a total count output of 8; one red + (one red+one green+ one blue) + (one red+one green+ one blue) + one blue = 8.

Therefore the relative luminance noise for the RWWB array is much smaller (square root of 2 smaller) & is therefore more desirable.


In essence, more photons are used to measure luminance the RWWB way. I don't know why this trick isn't implemented to boost S/N ratio.

Iowa Dave
You could ask Kodak.......
Color filter array - Wikipedia, the free encyclopedia
RGBW and RGBW 1,2, qnd 3
To the OP... check out this.....
GUILLERMO LUIJK >> TUTORIALS >> UNIWB. MAKE CAMERA DISPLAY RELIABLE

Last edited by jeffkrol; 06-06-2009 at 06:01 PM.
06-07-2009, 09:21 PM   #21
Veteran Member




Join Date: Nov 2006
Location: Hong Kong
Posts: 1,934
There is no White sensor on Earth!

QuoteOriginally posted by newarts Quote
It seems to me a color array of blue-white-white-red might be an improvement over the Bayer red-green-green-blue in use.

The RGGB array makes white by sum. A RWWB array would make green by difference (2G=W+W-3R-3B).
Unfortunately, AFAIK, there is no W photo diode or transistor sensor on Earth.

Even though there is filter in front of the sensing pixel, each of the R, G or B still has its own different peak spectral response on the visible spectrum. So, how can we make such a "W" wavelength response that has a flat response over the visible spectrum or has 3 peaks? Well, unless you put back three different photo diodes/transistors but then here we go back again!

(Please correct me if I am outdated for my knowledge and point me to the link of such a "White" pixel sensor)
06-08-2009, 02:58 AM   #22
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,863
QuoteOriginally posted by newarts Quote
It seems to me a color array of blue-white-white-red might be an improvement over the Bayer red-green-green-blue in use.
Dave, what a great suggestion and subsequent discussion!

Even if Kodak may have already had the idea

Your idea would decrease luminance noise and increase color noise, as you pointed out yourself. (color noise is increased because of error propagation in subtracting two noisy channels)

So, carrying your idea further, the best sensor may be one with two smaller unfiltered cells and two larger cells, filtered with secondary colors each. Problem is how to make cells unequal size ...
06-08-2009, 07:43 AM   #23
Inactive Account




Join Date: Dec 2008
Location: Ames, Iowa, USA
Photos: Albums
Posts: 2,965
QuoteOriginally posted by RiceHigh Quote
Unfortunately, AFAIK, there is no W photo diode or transistor sensor on Earth.

Even though there is filter in front of the sensing pixel, each of the R, G or B still has its own different peak spectral response on the visible spectrum. So, how can we make such a "W" wavelength response that has a flat response over the visible spectrum or has 3 peaks? Well, unless you put back three different photo diodes/transistors but then here we go back again!
I've already apologized for my sloppy use of language; read "White filter" as "clear filter" or "unfiltered".

Unfiltered light is red+green+blue. These separable quantities obey the laws of arithmetic; green = unfiltered - red - blue (so long as the underlying sensor characteristics are reasonable.)

Response to unfiltered light need not be "flat" RE wavelength for this to work.

Iowa Dave

06-08-2009, 07:56 AM   #24
Inactive Account




Join Date: Dec 2008
Location: Ames, Iowa, USA
Photos: Albums
Posts: 2,965
QuoteOriginally posted by RiceHigh Quote
Unfortunately, AFAIK, there is no W photo diode or transistor sensor on Earth.

Even though there is filter in front of the sensing pixel, each of the R, G or B still has its own different peak spectral response on the visible spectrum. So, how can we make such a "W" wavelength response that has a flat response over the visible spectrum or has 3 peaks? Well, unless you put back three different photo diodes/transistors but then here we go back again!
I've already apologized for my sloppy use of language; read "White filter" as "clear filter" or "unfiltered".

Unfiltered light is red+green+blue. These separable quantities obey the laws of arithmetic; green = unfiltered - red - blue (so long as the underlying sensor characteristics are reasonable.)

Response to unfiltered light need not be "flat" RE wavelength for this to work.

Iowa Dave
06-08-2009, 05:13 PM   #25
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
QuoteOriginally posted by falconeye Quote
Dave, what a great suggestion and subsequent discussion!

Even if Kodak may have already had the idea

Your idea would decrease luminance noise and increase color noise, as you pointed out yourself. (color noise is increased because of error propagation in subtracting two noisy channels)

So, carrying your idea further, the best sensor may be one with two smaller unfiltered cells and two larger cells, filtered with secondary colors each. Problem is how to make cells unequal size ...
you do realize this is like the Fuji sensor, Pretty sure R-pixel is unfiltered...

06-08-2009, 07:10 PM   #26
Veteran Member




Join Date: Nov 2006
Location: Hong Kong
Posts: 1,934
Not Viable, IMHO.

QuoteOriginally posted by newarts Quote
I've already apologized for my sloppy use of language; read "White filter" as "clear filter" or "unfiltered".

Unfiltered light is red+green+blue. These separable quantities obey the laws of arithmetic; green = unfiltered - red - blue (so long as the underlying sensor characteristics are reasonable.)

Response to unfiltered light need not be "flat" RE wavelength for this to work.

Iowa Dave
As I pointed out last time, the problem of a white (or "clear", no matter what it is called) pixel sensor is that the spectral response relies totally on the sensor itself as there is no filtering. But then no electronic component so far can detect the visible spectrum so widely from R to B. In fact, a typical sensor has a peak spectral response at G.

So, if such a clear sensor is to be made, the mostly viable choice is to *still* put in a "green" sensor. But then without cutting the blue and the red with the green filter (which you have picked away), the colour calculation will become rather difficult and *accurate* re-construction of original colours are almost impossible.
06-08-2009, 07:50 PM   #27
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
QuoteOriginally posted by RiceHigh Quote
As I pointed out last time, the problem of a white (or "clear", no matter what it is called) pixel sensor is that the spectral response relies totally on the sensor itself as there is no filtering. But then no electronic component so far can detect the visible spectrum so widely from R to B. In fact, a typical sensor has a peak spectral response at G.

So, if such a clear sensor is to be made, the mostly viable choice is to *still* put in a "green" sensor. But then without cutting the blue and the red with the green filter (which you have picked away), the colour calculation will become rather difficult and *accurate* re-construction of original colours are almost impossible.
think you should research a Foveon sensor, where the silicon itself acts as a filter.
Foveon - Direct Image Sensors
There is a filter layer ( I believe yellow so blue doesn't penetrate any farther ,just like film btw). The lower the wavelength, the deeper it penetrates the silicon....
The layers of pixels are embedded in silicon to take advantage of the fact that red, green, and blue light penetrate silicon to different depths – forming the first and only image sensor that captures full color at every point in the captured image.
Unfortunately it takes a lot of processing power to overcome the "uneveness" of the penetration and the lack of abrupt cutoff......

There is no CFA (except the yellow one between the blue layer and "green/red" layer).
Foveon's vertical color filter
Foveon's scheme of vertical filtering for color sensing

Another type of color separation, sometimes called vertical color filter, as opposed to lateral color filter, is used in the Foveon X3 sensor. Each location the image sensor array has three stacked active pixel sensors, utilizing the sensor's silicon itself as the color filter, based on the wavelength-dependent absorption of light in silicon.[8][9]

Bayer filter - Wikipedia, the free encyclopedia

And again, Kodak has researched CFA's for what, 33 years. IF a C,R,B sensor worked we would already know........
http://www.nabble.com/Kodak%27s-alternative-to-Bayer-sensor-pattern-td11122108.html
Some of the math involved in building an image from the matrix...
http://www.siliconimaging.com/RGB%20Bayer.htm
Just for curiosity:
http://www.patentstorm.us/patents/5374956/description.html

Last edited by jeffkrol; 06-08-2009 at 08:12 PM.
06-09-2009, 04:31 AM   #28
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
FRANKENSTEIN Sensor

Eureka... maybe.
After thinking about this awhile I came up w/ an interesting idea (feel free to tear into it)
Picture this:
A 15 effective megapixel sensor based on a foveon/kodak pattern hybrid sensor (think I'll send it off to Foveon today).
This 15mp based on the foveon layer technology (45mp of sensels). On top of this sensor you put the clear/red/blue filter array.
Now w/ the filtered pixels (stacked 3 deep) you gang the output, making them bayer like. The clear stacked pixel you can use not only the ganged output but you now have the possibility of all 3 channel data. Therefore it can function as a luminous pixel AND contain color data for demosaicing... get it!!
Sorry I'm not about to figure out how you use the data though.
One thing is (since it's CMOS) you can read out each pixel separately giving the possibility of having a 5mp b&w only readout (using the ganged "clear" output).
Anyone care to give me arguements for or against it???
The sensor hardware would have to be able to handle a 25mp (5x3 plus 10 )"sensel" output.
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
balance, camera, color, dslr, exposure, filter, image, lens, light, photography, quality
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
The Economy Is Improving graphicgr8s General Talk 22 05-21-2010 07:26 PM
K7 image autorotation driving me crazy! Spock Pentax DSLR Discussion 22 10-14-2009 08:47 AM
Improving AF with some MF finetuning? danielchtong Pentax DSLR Discussion 2 05-03-2009 01:04 AM
improving bokeh Urmas R. Pentax SLR Lens Discussion 35 04-08-2009 06:31 AM
My wife isn't crazy about this image... FHPhotographer Post Your Photos! 16 09-11-2008 03:29 PM



All times are GMT -7. The time now is 05:46 PM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top