Then the sensor has some number of pixels which are selective to the polarization, so that they get signal from only the left or the right half of the lens. Here are the examples of polarization selective pixel placement (Wv and Wh stands for white pixel sensitive to vertical and horizontal polarization, respectively):
Then one can determine the distance from seeing the parallax between the image sensed by W, Wh and Wv pixels - a nice way to implement a passive 3D camera.
Is this somehow connected to what Mukul Sarkar, student of Albert Theuwissen has done in his PhD related to insect inspired image sensors?
ReplyDeletepolarization sensitive pixel has a long history...
ReplyDelete@ "polarization sensitive pixel has a long history..."
ReplyDeleteTrue, but so far I have not seen any mass market application for such pixels. Panasonic application is the first one that can sell in hundreds of millions.
with the size of mobile phone CCM, there will be no 3D effect at all except you shot an object at 3cm! Classically in old style photography, the size of a lens is large and this gives quit large baseline for 3D sensation generation. You can find such stuff in popular science magasins. Panasonic is rich and can file as many gabage patent applications as they like.
ReplyDeleteI agree, this idea probably does not make sense in mobile phone camera. However, thinking about DSLR or high-end compacts, it might work well. More than that, it's hard to imagine a better solution for DSLRs, if 3D imaging becomes widely spread in compact cameras and mobile phones and DSLR vendors would have to respond.
ReplyDeleteHow do they do polarization on the pixel?
ReplyDeleteCould this also be used to provide speedy autofocus in a mirrorless camera?
ReplyDeleteI think the reason this wouldn't be useful in a mobile phone is that the lens polarizer will lose 70% of the light. Little sensors have a big problem with SNR, and this makes it substantially worse.
ReplyDeleteSince most applications I'm familiar with would like better SNR, this idea seems kind of pointless.
I don't know much about the market for single-lens cameras that measure 3D (cheaper Kinect?), but the usual reason to measure distance is to control an autofocus system. For this use, polarization is a fine idea, since the precision needed for autofocus scales with the precision available from any kind of stereo-inside-the-lens.
Maybe you could combine this idea with some kind of liquid crystal polarizer at the aperture of the lens, so that you could apply polarization to measure distance and control focus, and then take the shot with the polarizer off to get better SNR. You might be able to use the same polarizer as a global shutter (or half shutter) -- it'd be nice to get that functionality out of the CMOS pixel.