Bye bye Bayer? Panasonic claims new sensor tech ends color filter light loss


posted Tuesday, February 5, 2013 at 2:11 AM EDT

With the exception of Fujifilm cameras based around EXR and X-Trans image sensors and the Foveon-based camera lineup from Sigma, almost every digital camera on the market today shares one important feature in common: a Bayer color filter array, named after inventor and Kodak scientist Bryce Bayer. In fact, the overwhelming majority of color cameras made since the very advent of digital photography have been based around Bayer filters. If research work by Panasonic Corp. comes to fruition, though, that might well be changing in the not-too-distant future.

A couple of days ago, the peer-reviewed scientific journal Nature Photonics published an article coauthored by four staff members of Panasonic's Device and Cloud Solutions Centers in Osaka, as well as one individual from Panasonic's Image Sensor Business Unit in Kyoto. A press release from the company briefly summarizes the article's contents, and it promises good things to come for photographers -- that is, if the technology can be successfully commercialized, and proves not to have significant drawbacks of its own.

The basic goal of the research, which Panasonic says has garnered 21 patents or patent applications in its home market and another 16 overseas, is fairly simple. It recognizes that while Bayer filters have allowed the recreation of color images from purely monochromatic image sensors, they have a significant drawback. They work by absorbing -- and thereby, blocking -- incoming light at each photodiode, allowing only certain ranges of wavelengths to pass. In the process, a very significant portion of the incoming light is blocked before it can reach the sensor.

Panasonic's research aims to replace the color filter array with a more efficient alternative that wastes less of the incoming light. What the company describes is a layer of micro color splitters that replace the filter array, and are fabricated from inorganic materials using existing semiconductor fabrication techniques. The splitters can, says Panasonic, be used regardless of whether the underlying image sensor is based around CMOS or CCD technology. That's pretty important, because it means that should the research prove suitable for real-world use, it should be relatively straightforward for sensor fabs to implement.

A row of photodiodes from a standard Bayer-filtered sensor (left) compared to one using Panasonic's micro color splitter technology to replace the color filter array (right).

When incoming light reaches the splitters, it is separated as in the example on the right above. Half of the photodiodes in this example receive everything but light within the red wavelengths, or white minus red (W-R), as it's described in the diagram. The other half receive the incoming light for their own photodiode, plus half of the red light from each of two adjacent photodiodes. (That's white plus red, or W+R.) And that's only one row on the sensor; the diagram doesn't show this but the next row would similarly feature alternating white minus blue (W-B) pixels, and white plus blue (W+B) photodiodes.

Where the standard Bayer array blocks around 50-70% of the incoming light, according to Panasonic, the company says that its beam-splitter allows fully 100% of the incoming light to reach the sensor surface. That increase in captured light leads to a significant increase in the sensor's sensitivity, on the order of two to threefold according to the press release. The abstract of the article on Nature Photonics suggest that early prototypes show somewhere on the order of a 1.85x increase in light captured, which would still be a huge step forward. In the interests of simplicity, imagine a twofold increase in light captured: that would in theory yield double the sensitivity for a given noise level. (Or looked at the other way around, half the noise level for a given sensitivity.) All else being equal, a properly-exposed shot at ISO 3,200 captured with a micro color splitter-based sensor would have the same noise levels as one shot at ISO 1,600 with your existing camera.

Of course, that's just the theory. In practice, it remains to be seen if the technology can be commercialized, and if so what drawbacks it might bring. There are doubtless some complex issues to be resolved, and it isn't yet clear how close Panasonic has come to addressing these. In particular, we're curious about how the micro color splitters handle the increasing angle of incidence of incoming light towards the corners of the sensor: A supplementary-info PDF seems to suggest some pretty nasty behaviour changes with increasing angle of incidence. (Figures S5a and S5b in that document.) Saturation of photodiodes might also be a challenge, given that some would now be responsible for gathering light not only for their own location, but also for those of adjacent photodiodes.

Two images shot with CCD image sensors having equal sensitivity. At left, the sensor is overlaid with a Bayer filter. At right, it is said to use Panasonic's micro color splitters.

And then there's the issue of the processing required to recreate the color information for the image, with information from multiple pixel locations having been combined into one. That may be critical not only in terms of processor power required -- it would appear rather less straightforward than demosaicing Bayer data where each photodiode is a distinct entity -- but also in terms of the potential for reduced resolution and increased noise levels in that color information.

Still, Panasonic is clearly hard at work on the technology, and surely sees a (bright) light at the end of the tunnel if it can surmount the obstacles in its path. With a patented technology allowing a dramatic improvement in noise / sensitivity in its hand, the company could hold quite a trump card over its rivals -- and a clear advantage in the eyes of photographers. And while we have a great deal to thank Mr. Bayer for, after a decade-plus of public and press alike using his invention to document the world around us, we're more than a little excited by the prospect of that new generation of cameras, should Panasonic hit paydirt!

For those of a technical bent, the Panasonic article is available to read on the Nature Publishing Group website, should you have a subscription or care to buy access. (There's also a PDF file containing supplementary information available as a free download.)