New sensor from Kodak, a hint of technology Panasonic might look into someday?

marco0782

marco0782

Well-known member
It's a 1/4" 5-megapixel CMOS sensor that's designed for mobile phones and other small applications but using a new technology developed by Kodak (clear pixels in addition to RGB pixels) it achieves 3200 ISO which is incredible for such a small sensor.

Light sensitivity in the new sensor is enhanced through the use of the recently announced KODAK TRUESENSE Color Filter Pattern, which adds panchromatic, or “clear,” pixels to the red, green and blue pixels already on the sensor. Since these pixels are sensitive to all wavelengths of visible light, they collect a significantly higher proportion of the light striking the sensor. This provides a 2x to 4x increase in sensitivity to light (from one to two photographic stops) compared to current sensor designs, improving performance in low light and reducing motion blur in action shots.
Click to expand...

The whole article is available here.
 
I'm afraid you never get anything for nothing.

Adding the clear sensors does help the sensitivity - but at the expense of chroma resolution. Hence it's got merit for such as mobile phone cameras, but not for the sort that people on this board are most interested in.

And a sensor can have whatever ISO rating you wish it to have, it just depends how much gain, noise and processing you're prepared to put up with. For a mobile phone camera that's a lot more than a more conventional camera.
 
The idea is that it achieves 3200 without gain or noise, obviously. And this chip is capable of 720p at 30fps so it is not only for stills.

I would like to know how the clear pixels are at the expense of chroma resolution. It's smart for Kodak to start off with mobile phones as they develop this new technology. I bet that it will eventually be very desirable in other applications.


Marco
 
marco0782 said:
The idea is that it achieves 3200 without gain or noise, obviously.
Click to expand...
The exact wording of the press release is "....enables imagery up to ISO 3200 and support for full 720p video at 30 fps", and I think the words "up to" are pretty relevant. A digital stills camera can be whatever ISO rating you dial into the menu, higher settings obviously being grainier and noisier than lower. "Up to" in this context implies that at this sensitivity it will be far from noise free.
I would like to know how the clear pixels are at the expense of chroma resolution.
Click to expand...
Because only filtered pixels can give information about the colour of a particular area, through an R,G, or B filter over each one, which will absorb the other two primary colours. Hence light is inevitably lost in the filtration process.

Take away the filters from 1 each R,G,B pixels, make three clear pixels and ALL the light then reaches those photosites - so an immediate sensitivity improvement. But then we no longer know whether that area of the picture is red, green, blue or whatever. It may be guessed at by interpolation from the coloured sensors left - but there are now less colour pixels to use for the interpolation (compared to a Bayer array) hence lower chroma resolution. You don't get something for nothing.

The press release does say "...designed for mass-consumer camera applications such as mobile phones...." which implies small size being important, certainly more so than absolute quality. It's a valid and clever technique for such devices, but less relevant for dedicated cameras, where absolute quality IS important, size less so.
 
The way I understand it, is that 4th pixel is the brightness and the other three are the color pixels, then all four are summed into one full color pixel...... so the sensor is only capable of 1/4 of the total pixels.... so the sensor would be effectively 1.25 megapixels.
 
Last edited:
Actually, it's just like having chroma subsampling built right into the chip. Each pixel has luminance data, so the black and white is full res. the color is 1/4 res. Formats like MiniDV discard 75% of the color information anyway.

You have luminance plus RGB (kind of like reverse CMYK). If you could get the RAW data you could have some really cool options in post. Not that anyone would build an andromeda for their cell phone.
 
Andrew McCarrick said:
The way I understand it, is that 4th pixel is the brightness and the other three are the color pixels, then all four are summed into one full color pixel...... so the sensor is only capable of 1/4 of the total pixels.... so the sensor would be effectively 1.25 megapixels.
Click to expand...
Not quite, and whilst most references to the sensor just repeat the press release pretty much verbatim - this one: http://www.imaging-resource.com/NEWS/1181811769.html gives much more information - they seem to be typically talking about half the total pixels being clear, quarter green, and 1/8th each red and blue. The announcement seems to concern a concept, not the exact pattern. It's worth repeating one quote from that link:
Depending on the application, different patterns may be more appropriate for use. For example, one natural trade-off is the balance between the sensor’s overall sensitivity (via the pan pixels) and how well the sensor collects color information (via the RGB pixels). The highest sensitivity would come from a sensor composed only of pan pixels, but would provide no color information.
Click to expand...
In other words, chroma resolution can be traded for sensitivity, up to a point.

If it was as you said above, it would be difficult to give an exact figure for the resolution, but would be less than 5 megapixels, more than 1.25 - that's the effect of the De-Bayering algorithm. In practice it's likely to tail off, so there just is not an exact figure. That's for luminance, the chrominance figure will be less, and again it's difficult to accurately quantify.
 
You must log in or register to reply here.
Back
Top