facebook rss twitter

Canon 35mm full-frame CMOS sensor excels in low light

by Mark Tyson on 4 March 2013, 17:00

Tags: Canon (TYO:7751)

Quick Link: HEXUS.net/qabthn

Add to My Vault: x

Canon Inc. has some interesting news today regarding a new photographic sensor in development. The Japanese imaging giant has “successfully developed a high-sensitivity 35mm full-frame CMOS sensor exclusively for video recording”. It’s very sensitive and provides low-noise full HD video capture in “exceptionally low-light environments” where the naked eye would find it hard to perceive objects.

How it works

Canon uses a 35mm full-frame CMOS sensor with pixels measuring 19 microns square. These, the company informs us, are over seven times larger than the pixels on the top-of-the-line EOS-1D X camera. Additionally there are new technologies employed to reduce noise.

Capabilities

The easiest way to see the new senor’s capabilities are by watching the video above. If you are interested, the version of this video on Canon’s site seems a bit better quality than the YouTube one I could embed above, which has probably been recompressed.

Canon’s press release says that the sensor “facilitates the shooting of clearly visible video images even in dimly lit environments with as little as 0.03 lux of illumination, or approximately the brightness of a crescent moon—a level of brightness in which it is difficult for the naked eye to perceive objects”. A video recording made with an electron-multiplying CCD (EMCCD) shows about as much detail as the human eye can see, capturing magnitude-6 stars. The new CMOS sensor by Canon can capture much fainter images; stars with a magnitude of 8.5 and above.

Also in the video you will see a subject in a dark room with his face lit by just the light from a burning incense stick and footage of meteors from the Geminid meteor shower. I was particularly impressed by the video taken under a full moon. Canon showed a normal lifelike scene at night but also showed what the scene could look like with the sensitivity control increased; the scene was very light and detailed.

As well as nature and astronomy applications the sensor would also probably be very useful in night-time security video recording. If you are in Tokyo tomorrow (or any time until the 8th March) you can pop into the Security Show 2013 at Tokyo International Exhibition Center and see a prototype of the camera and some sample footage.



HEXUS Forums :: 4 Comments

Login with Forum Account

Don't have an account? Register today!
Basically its a 35mm 2 megapixel camera.

The other thing is they are showing video with this, which maybe because I don't do video really, makes me curious that they are doing something cool in time domain. Because with video you can observe each pixel for longer, potentially use some clever noise removal logic to clean up the image even more.

This is important because the lower the light, the more noise creaps in.

I would be interested to see a side by side still comparison from this sensor with something like a downsampled D800, as after all you could take 16 pixels and turn it in to one, a great way to have less noise.
TheAnimus
Basically its a 35mm 2 megapixel camera.

The other thing is they are showing video with this, which maybe because I don't do video really, makes me curious that they are doing something cool in time domain. Because with video you can observe each pixel for longer, potentially use some clever noise removal logic to clean up the image even more.

This is important because the lower the light, the more noise creaps in.

I would be interested to see a side by side still comparison from this sensor with something like a downsampled D800, as after all you could take 16 pixels and turn it in to one, a great way to have less noise.

IMO the reason they used video is because there is no expectation of high resolutions. 1080p is a good video resolution that has people going ‘ooh shiny’ whereas a 2MPx stills camera is lolworthy. Obviously at 30fps you are limited to minimum shutter speeds of 1/30th of a second - however I don't see this as an issue since longer than that and unless you're imaging stationary objects, you'll get motion blur anyway.

I'm not convinced by the noise reduction logic either - it's been possible for years to take two exposures of the same (stationary) object for noise reduction purposes. With video there is a presumption that you are recording moving objects. So even if you reduced the frame rate by half in aid of noise reduction, you would still be trying to reduce noise from two pictures that are not identical, so would be left with a blurry image.

Comparing this to downsampling I would expect this to be a little better since you would not have the dead space between photon detector sites present in downsampling. However the other argument is that by downsampling you are also performing noise reduction at the same time, so I'm not sure what the end result would be.

Pretty exciting stuff though :)
Another technique used by still cameras is dark frame subtraction, after recording the image, the shutter is closed and another image of equal exposure time is taken, and subtracted from the original. Some noise tends to remain fairly constant, at least between two successive shots. This is partly why a DSLR set to expose for 10s will spend another 10s ‘processing’.
After been able to watch the video its interesting that they are out performing NEC's secuirty oreintated 3 sensor camera, which relies on a prisim to break the light in too three paths for the red green and the blue, this reduces the amount of light lost in the filter.

However they are beating that. Which in a way is quite impressive, however I do feel its a bit like looking at the fastest steam boat, in the age of the airplane.