Indeed, the technology of 3D structures is catching up with some very old ideas going back at least to the 70's. The NASA Massively Parallel Processor was proposed in 70's at Goddard (I think). Hughes (e.g. Graham Nudd) did a lot of work on stacked photodiodes and processor planes. Most of this early stuff is referenced in my PhD thesis from 1984 at Yale. In that thesis I was working on discrete time CCD analog processors for a focal-plane image processor, and even have a cartoon of an array of photodiodes stacked on a parallel array of processors. My student at Columbia, Sayed Eid, continued this work and built a focal-plane image processor using CCD (charge packet) analog computer elements which was published in some SPIE proceedings. We were not the only ones looking at focal-plane image processing. I remember work by Paul Beaudet at Westinghouse and a group at Honeywell, and some other related work by Alice Chiang at MIT/LL. At some point in the latter 80's, roughly coincident with my moving to JPL, I decided it was better to put the ADC on the image sensor and do the focal-plane processing in the digital domain (hence the theses of Sabrina Kemeny and Suni Mendis). Then we went off to CMOS image sensor land and I figured other people were better at digital image processing than I would ever be, and focused on building the CIS platform technology. Stacked BSI CIS has been the holy grail platform technology for focal-plane image processing for 40 or 50 years. Glad to see Watanabe exploring and demonstrating this sort of technology now. I am guessing there are others doing the same around the world.
This is more interesting than it may at first appear; or it's unworthy of a comment.
ReplyDeleteThis Image Sensor has limited intelligence but in the future it could provide a 'seeing device' for the Blind; a 'Smart Cane'.
The current ability to hit a baseball is excellent.
Indeed, the technology of 3D structures is catching up with some very old ideas going back at least to the 70's. The NASA Massively Parallel Processor was proposed in 70's at Goddard (I think). Hughes (e.g. Graham Nudd) did a lot of work on stacked photodiodes and processor planes. Most of this early stuff is referenced in my PhD thesis from 1984 at Yale. In that thesis I was working on discrete time CCD analog processors for a focal-plane image processor, and even have a cartoon of an array of photodiodes stacked on a parallel array of processors. My student at Columbia, Sayed Eid, continued this work and built a focal-plane image processor using CCD (charge packet) analog computer elements which was published in some SPIE proceedings. We were not the only ones looking at focal-plane image processing. I remember work by Paul Beaudet at Westinghouse and a group at Honeywell, and some other related work by Alice Chiang at MIT/LL. At some point in the latter 80's, roughly coincident with my moving to JPL, I decided it was better to put the ADC on the image sensor and do the focal-plane processing in the digital domain (hence the theses of Sabrina Kemeny and Suni Mendis). Then we went off to CMOS image sensor land and I figured other people were better at digital image processing than I would ever be, and focused on building the CIS platform technology. Stacked BSI CIS has been the holy grail platform technology for focal-plane image processing for 40 or 50 years. Glad to see Watanabe exploring and demonstrating this sort of technology now. I am guessing there are others doing the same around the world.
DeletePut some good & stable framework. Release a good open source language and be amazed what can be achieved towards intelligence. ;-)
ReplyDelete