Quote from: wattMaster on May 14, 2016, 08:47:37 AM

I would think that high-end cameras would be calibrated/designed to more accurately take pictures like the eye.
I think it comes down to the color filters cameras use, There are really only 2 common filters, Bayer and Fujifilm.
The Fujifilm filter boasts to have better colors, And more accurate rendering.

Well, that works if there are filters.
The thing is, the color borders with the camera filters tend to be rather sharp cutting, but in the eye they are way smoother and that is quite hard to emulate.
But mainly the new CMOS sensors do not use any explicit color filters at all (except to block the off-visible wavelengths). Their sensor elements use stacked junctions in different depth (Nplus/Pwell/DeepNWell/Psubstrate; before the shutter window begins, they are precharged to a reference voltage so all are in reverse, then when the shutter window starts, the precharge switches are switched OFF, so the light created current discharges the junction capacitances, after the exposure the voltages are measured via an ADC, then corrected for the capacitance nonlinearity), each sensing different parts of the spectrum (and their spectral sensitivity overlaps). Then the resulting signal is recalculated for the standardized RGB values.
The reason is to suffice with the most common CMOS manufacturing process without any added steps (all the layers are already present, originally designed for the basic NMOS and PMOS transistors). The other benefit is, such concept allows the complete pixel area to sense all color components, so one pixel is really a full color pixel, no smaller subpixels or so.
And here is not much room for physical adjustments, so the color recalculation matrix is just optimized for the most common scenes (that is, why there different scene settings even when the CCT and exposure settings are identical - it is to use a matrix better optimized for different light sources)

Do you know of any cameras that use this technology?
And here's fujifilm's filter: http://www.fujifilm.com/products/digital_cameras/x/fujifilm_x_t1/features/page_02.html
It looks like the micro lenses make the image smoother.

Do you know of any cameras that use this technology?

It was no exact camera model, but and advertisement for a picture sensor IP (so the company who designed it sells mainly the licenses, to be integrated by the final sensor IC maker; very similar business model as ARM is using for their processors). It was describing the exact pixel structure cross section details and presenting mainly the benefits of that concept (and one of the benefits is the nonexistent color moire; the color accuracy they obviously didn't mention explicitly). The article was written so, it assumes the basic pixel sensor structure as the generic knowledge, used as one of the most common methods. Theirs was then the complete array design and the associated control and sensing electronic (and probably the color correction tables), likely in the form of some sort of "compiler" (you set in the basic parameters, like resolution in each direction and pixel size) and the compiler then generates the complete gds view (so the layout of all the layers in the target fabrication process) and simulation macro models (so you, as the final chip designer are able to verify all the complete IC will likely work well before even ordering the masks for 100+k$).

I would guess the use could be either for some lower cost cameras (I would guess for the mobile phones or so), or some special picture sensors (likely industrial machine vision systems, where the integration with special functionality was needed, hence as well the stressed no need for extra steps).

It was aimed for some TSMC 110nm node; these technologies are used by many IC companies (they design their own chip, but for fabrication use some foundry vendor technology - TSMC is one of the biggest such foundry service vendors)...

And here's fujifilm's filter: http://www.fujifilm.com/products/digital_cameras/x/fujifilm_x_t1/features/page_02.html
It looks like the micro lenses make the image smoother.

This looks more like some "higher end" sensor type, I doubt that would be used in any cost sensitive device like mobile phones or so.

Quote from: wattMaster on May 15, 2016, 08:05:35 AM

And here's fujifilm's filter: http://www.fujifilm.com/products/digital_cameras/x/fujifilm_x_t1/features/page_02.html
It looks like the micro lenses make the image smoother.

This looks more like some "higher end" sensor type, I doubt that would be used in any cost sensitive device like mobile phones or so.

And, I guess it could be used in specialized equipment to measure CRI.
What about the handheld CRI meters that use image sensors like these?
Or how about a way to decipher images of light sources and get the CRI from them?
(Take a photo of the thing you want to measure, Upload it into the program, Then select what camera you have.)

The only thing the camera can get by itself is the light color (the color coordinates; more or less accurately, but that is not important), that is not the CRI.
The CRI is the ability of a light source to render colors of a colored object "accurately" (that means the same way as the ideal blackbody radiator of the same intensity and color temperature would do). So what matters is not that much the color of the light itself, but if the color and intensity of the light reflected from the colored object is the same as when it is illuminated by the reference blackbody radiation. In other words how that light distorts colors of the colored objects.

The color coordinates and color rendering are two very different parameters, describing very different aspects.

E.g. an incandescent with a green lacquer will emit green light, so it's color will be very off-white. But because it is continuous spectrum, it will still all colors are the same as with the naked incandescent.
Compare to that you may have a DX MV lamp. Perfectly white light by itself (and a really accurate sensor will pick it white as well), but all red objects  turns brown. So color is pretty accurate white (way way closer than the green lacquer incandescent), but the color renderng is significantly worse (way worse than with the incandescent example before)

So the RGB sensor can not pick up enough data to judge the CRI, you need really spectrum analysis for that. And so you need either sensor with each pixel having narrow band filter for a given spectrum line (there are sensors doing that, no idea what is the filter system there, but it is clearly directly on the sensor chip), or you have to use some optical method to redistribute the light according to the wavelengths (diffraction grating, e.g. the CD, and/or a prism). Then when the resulting picture is sensed by a color sensor, the RGB mix may help to calibrate the thing (each wavelength has certain color and that could be sensed by the RGB ratio - when you know the light is of just a single spectral line, the resulting RGB ratio gives you quite accurately it's exact position)

Good news, I just got an email, And the CRI is 71.