Sensor manufacturers are constantly working to increase the signal to noise performance, dynamic range and color reproduction quality. If you have more technical skills you may check out the two new Sony patents showing two ways to realign the single RGB pixels to increase the image quality: Patent one and patent two. In the past we saw other very “alternative” patents (like foveon alike pixel alignment and so on). From what I am hearing from my sources some of this paradigm changing sensors will see the light of day in a real camera by 2014.
18 Responses
2014…well…hope we’ll all live long enough to see that…. 😉
But imagine if this comes truth – we’ll have a new record of “foreseeing rumors” on SAR. 😉
Soon more “reliable” 2013 rumors coming too 😉
hehe, and how am I suppose to break with my addiction to SAR? I can’t… I just can’t… not with things like that coming in…. 🙂
What happened to all the organic sensor hype of last year?
I guess nothing – patents for stuff like that usually are made years before actual implementation in commercial products.
Patent 1 is for reducing colour shading in the periphery of the sensor. Still uses a conventional Bayer layout (albeit slightly shifted to even out colour response).
Patent 2 still uses a conventional Bayer layout, but the effective aperture of the pixels can be different. One set of pixels can respond to the incoming light of a large aperture lens. The other set of pixels responds to a narrower cone of light, hence take an image with a smaller effective aperture (greater depth of field). Images are reconstructed from both sets and then combined to optimise both in-focus and out-of-focus areas. One can see it as a distant variant of plenoptic imaging.
Your description of the second patent blew my mind. Now I’m reading about light-field cameras (like the Lytro) on Wikipedia, hmm…
Good summary, thanks! In patent 2, it looks like the specific objective is to combine the shallow depth of field of a wide aperture with the greater in-focus sharpness of a narrow aperture. So your picture might have the depth of field of f2, but the areas in focus would be as sharp as f8.
Hope Sony give the RAW converters a 1 year headstart .
😉
Well, that actually is good point considering RAW support for either Fuji or Sigma. But this patter RGrGbB is quite easy to demoisaic as its same pattern as original RGGB and should require just color rebalancing/recalculating.
Tho its still wrong patter, theres possibility of better one. I just cant figure why no one figured it yet. Maybe I should patent it. 😀
Red, Green, Blue and Orange would be best if one has to stick with groups of four pixels.
This would match up better with the latest inkjet printers, which use more than four ink colours, including an orange.
It’s just a bit more complicated than that. For accurate colour (independent of the output device), you want the spectral response of the pixels to satisfy as much as possible Luther’s condition, i.e. a linear combination of the S, M, and L spectral response functions of the cones. However, you have to also consider the spatial domain response of the colour array pattern. To optimise both colour sensitivity and resolution without artificts, there’s some interesting research on working in the combined spatio-spectral regime. Look at http://campus.udayton.edu/~ISSL/index.php/research/optimal-color-filter-array/ for some articles on this.
On an RGB color wheel, Orange is complimentary to… Emerald! (Gb)
As is
Red: Cyan
Green: Magenta
Blue: Yellow
Looks like folks are on the same page of correcting for accurate color perception.
However, until RAW converters and PS start carrying Gb channels, everything will go through a RGB conversion. (Or if a printer can read a RGGbB RAW file directly…)
Didn’t Fuji make some sensors with two slightly different color G sensors in each array? A bit like the Gb sensor here.
Id this intended for small sensors used in phones or P&S cameras – or is it for larger sensors (APS-C and FF)?
It was actually Sony, http://www.imaging-resource.com/NEWS/1058329666.html
“Sony plans to integrate the newly developed 4 color filter CCD and the new image processor into the new digital cameras which will be introduced in the near future”. – SONY, 2003 lol.
It was used in the F828 http://www.dpreview.com/reviews/sonydscf828
I should point out that the Foveon does not use a pixel pattern. There is no pixel pattern used because its a stacked sensor. It doesn’t require demosiacing since every pixel has the full colour detail in it. There is no antialiasing required because there is no Bayer or X-Trans like array (aka pixel arrangement).
Foveon’s sensor is the way to do it – Its too bad Sigma bought Foveon – because Sigma can’t make a camera to save their life. If Nikon or Canon had acquired Foveon, I’d be headed over to their camp.