CNET: Keith Fife and his colleagues from Stanford University presented a multi-aperture image sensor on the recent ISSCC. Fife's 3MP prototype divides the pixel array into slightly overlapping 16x16-pixel patches called subarrays. Each subarray has one common microlens - thus the term multi-aperture.
The image differences from one subarray to the next can be used to extract the distance information from the image. Keith also claims other potential advantages, like improved noise filtering.
I don't think this sensor has its market yet, but it's always interesting to see fresh ideas, very different from what others are doing. This is why I like Keith Fife's papers.
The multi-aperture idea on its own is not new. I've seen a spectacular demonstration of a 3D endoscope a few years back using a very similar multi-aperture idea on a standard CCD. Us patent 6704043 describes it.
ReplyDeleteBut Keith's implementation in CMOS and integration with the small pixels are indeed very nice.
Wow! Collective knowledge can make wonders! I've never heard about this idea, other than from Keith's and his co-workers papers and presentations, starting from about a year ago.
ReplyDeleteI think the prior art you posted would come as a surprise for Stanford guys as well.
By the way, Keith's ISSCC paper describes CCD too, even though his CCD is made in CMOS process. So, the difference between the prior art and the Stanford group's work is even smaller.