Patent: Canon RGBW 12 Image Sensor

[Canon Rumors]

A <a href="http://egami.blog.so-net.ne.jp/2016-12-16">patent over at Egami</a> showing Canon is working on an RGBW image sensor has appeared.</p>
<p>Patent publication number 2016-208093 (Google Translated)</p>
<ul>
<li>Release date 2016.12.8</li>
<li>Application date 2015.4.15</li>
<li>RGBW 12 array</li>
<li>Generate resolution data and color data, synthesize resolution data and color data, up-convert, process with mosaic (Bayer conversion), demosaic in order</li>
</ul>
<p>The new patent shows how to get increased sensitivity, but sacrifices color resolution.</p>

		<style type='text/css'>
			#gallery-1 {
				margin: auto;
			}
			#gallery-1 .gallery-item {
				float: left;
				margin-top: 10px;
				text-align: center;
				width: 50%;
			}
			#gallery-1 img {
				border: 2px solid #cfcfcf;
			}
			#gallery-1 .gallery-caption {
				margin-left: 0;
			}
			/* see gallery_shortcode() in wp-includes/media.php */
		</style>
		<div id='gallery-1' class='gallery galleryid-27706 gallery-columns-2 gallery-size-thumbnail'><dl class='gallery-item'>
			<dt class='gallery-icon portrait'>
				<a href='http://www.canonrumors.com/wp-content/uploads/2016/12/2016_208093_fig01.png'><img width="168" height="168" src="http://www.canonrumors.com/wp-content/uploads/2016/12/2016_208093_fig01-168x168.png" class="attachment-thumbnail size-thumbnail" alt="2016_208093_fig01" /></a>
			</dt></dl><dl class='gallery-item'>
			<dt class='gallery-icon portrait'>
				<a href='http://www.canonrumors.com/wp-content/uploads/2016/12/2016_208093_fig03b.png'><img width="123" height="140" src="http://www.canonrumors.com/wp-content/uploads/2016/12/2016_208093_fig03b.png" class="attachment-thumbnail size-thumbnail" alt="2016_208093_fig03b" /></a>
			</dt></dl><br style="clear: both" />
		</div>

<span id="pty_trigger"></span>

 

[BeenThere]

Maybe for use in low light surveillance where color info is of secondary importance?

 

[GMCPhotographics]

Maybe for use in low light surveillance where color info is of secondary importance?

Maybe, but in theory...a sensor with a higher native DR should be able to take a far higher post processed contrast curve than compared to a normal sensor Bayer RGB array.

 

[mnclayshooter]

I keep thinking dual-pixel AF was a stepping stone to an all-in-one mirrorless/AF/light sensitivity stacked Birthday/layer cake type sensor. Sort of a proof of concept and now let's add another dimension to it with white and potentially a deeper stack of flavor layers.

Once we no longer need the pentaprism, mirror and separate phase-detection sensor, form factor can quickly morph to something very workable for high-dynamic range mirrorless or even more compact.

I know nothing though. I'll leave it at that.

 

[mb66energy]

Maybe for use in low light surveillance where color info is of secondary importance?

Good remark, it's more important to see a burglar but sufficient to have some hints about the colors of his/her clothing.
Canon is, as far as I know, deeply into this market segment, maybe as a new column to stabilize losses in the photo segment.

 

[Fleetie]

Maybe for use in low light surveillance where color info is of secondary importance?

Good remark, it's more important to see a burglar but sufficient to have some hints about the colors of his/her clothing.
Canon is, as far as I know, deeply into this market segment, maybe as a new column to stabilize losses in the photo segment.

I can imagine it working very well for exactly that application.


Lovely, high-resolution, high-sensitivity luminance; vague, washy chrominance that "will do".

 

[Sharlin]

Maybe for use in low light surveillance where color info is of secondary importance?

Good remark, it's more important to see a burglar but sufficient to have some hints about the colors of his/her clothing.
Canon is, as far as I know, deeply into this market segment, maybe as a new column to stabilize losses in the photo segment.

I can imagine it working very well for exactly that application.

Lovely, high-resolution, high-sensitivity luminance; vague, washy chrominance that "will do".

How much perceived difference can there be, anyway, at high enough resolution? The human eye is pretty poor at discerning color contrast, compared to luminance contrast, a fact already exploited by JPEG and MPEG compression as well as "regular" Bayer mosaicing.

Anyway, another use case jumps readily to mind: astro!

 

[midluk]

The increased sensitivity of these white pixels will likely come at the cost of an increase of the minimum possible ISO number.
The white pixels should receive roughly three times as much light as the color pixels. Which means they saturate at three times lower light levels.

 

[TAF]

Is it possible to make a strictly B+W sensor that has extremely high sensitivity?

One without the Bayer filter (unnecessary), which also yields high resolution?

That might be an interesting option...a 5DIVbw? Or an EOS 5bw?

 

[DominoDude]

Is it possible to make a strictly B+W sensor that has extremely high sensitivity?

One without the Bayer filter (unnecessary), which also yields high resolution?

That might be an interesting option...a 5DIVbw? Or an EOS 5bw?

If we neglect the Sigma Foveon sensor, all sensors are basically photon counters that couldn't give a damn about the colour of incoming light. So, yes it is possible and have been done.

 

[Sharlin]

Is it possible to make a strictly B+W sensor that has extremely high sensitivity?

One without the Bayer filter (unnecessary), which also yields high resolution?

That might be an interesting option...a 5DIVbw? Or an EOS 5bw?

http://us.leica-camera.com/Photography/Leica-M/LEICA-M-MONOCHROM

Dedicated astro cameras are almost always monochrome; to get color images you'd want to use separate exposures with different filters anyway.

 

[Rudeofus]

The increased sensitivity of these white pixels will likely come at the cost of an increase of the minimum possible ISO number.
The white pixels should receive roughly three times as much light as the color pixels. Which means they saturate at three times lower light levels.

Your estimate assumes that bayer filters are perfect - and they are anything but. Not only is color separation rather poor, there is also significant dampening of light wavelengths which should actually pass through, to the tune of "only 1/4 of the blue light goes through the blue filter".

 

[scottkinfw]

Fascinating stuff, but above my pay grade.

Please school me. If this was for a specific, perhaps a niche application like night surveillance as above, that is likely good? For us mainly stills shooters, would this be a good thing? I am concerned when there is a seemingly backward step in resolution. I don't shoot with the 5DSR series, but still, I want the best resolution for my purposes.

Thanks for the insights.

Scott

 

[mb66energy]

The increased sensitivity of these white pixels will likely come at the cost of an increase of the minimum possible ISO number.
The white pixels should receive roughly three times as much light as the color pixels. Which means they saturate at three times lower light levels.

Your estimate assumes that bayer filters are perfect - and they are anything but. Not only is color separation rather poor, there is also significant dampening of light wavelengths which should actually pass through, to the tune of "only 1/4 of the blue light goes through the blue filter".

Spectral color separation must be poor to detect e.g. yellow - you need sensitivity of red and green photosites to add the signals together and make a valid color triplet. Spatial separation - that's true - is bad because of the lower count of red green and blue pixels compared to the luminosity "channel" of all pixels.

The low blue sensitivity is IMO caused by filter PLUS photodiode material sensitivity - so it is a matter of new photodiode materials and not about the filter dye.

 

[Stuart]

Apple is using 3 Sensors and a Prism That Splits Light - to save losing light in a Bayer filter. That seems neat.

I wonder when Fabs will be able to produce an array of tiny prisims that collect light and then split it to, at least, 3 different sensing locations at each microsite group.

 

[Rudeofus]

The low blue sensitivity is IMO caused by filter PLUS photodiode material sensitivity - so it is a matter of new photodiode materials and not about the filter dye.

That was my first thought when I saw the chart, but the label explicitly states "Bayer Filter Transmission Spectral Profiles". There is no other way to read this than that this chart shows just filter characteristic.

See also the text right above the chart:

Typical transmission spectral profiles of the common dyes utilized in the construction of Bayer filters are presented in Figure 4. The quantum efficiency of the red filters is significantly greater than that of the green and blue filters, which are close to each other in overall efficiency.

PS: photographic slides gained lots of brilliance when they switched from additive to subtractive dyes over 80 years ago. I wonder when digital sensors will follow this concept.

 

[TAF]

Apple is using 3 Sensors and a Prism That Splits Light - to save losing light in a Bayer filter. That seems neat.

I wonder when Fabs will be able to produce an array of tiny prisims that collect light and then split it to, at least, 3 different sensing locations at each microsite group.

Given a variation on DLP (digital light processing) technology, it seems like the state of the art would allow something like that now.