Cornell University Chronicle: Alyosha Molnar, assistant professor of electrical and computer engineering, and his group are building light field image sensors that give readouts of the incident angle of light as it strikes the sensor. The result could be the next generation of 3-D cameras, as well as the ability to focus photos after they're taken, similar to Lytro. This work appears to be a continuation of one presented at ISSCC 2011 (See: A. Wang, P. Gill, and A. Molnar, " An Angle-Sensitive CMOS Imager for Single-Sensor 3D Photography", also in Jan., 2012 issue of JSSC).
"
If I want to find something in 3-D space, if I just measure the amount of light hitting different locations, I know a little bit. I'll get some sort of blurry blob," Molnar said. "
But if I know the incident angle of the light, I can triangulate back. So the question is how do you measure that?"
Few slides from
Alyosha's group research site explain the principle:
The sensors have resolution of 150K pixels, but the increased resolution is promised. The latest sensors are implemented in 130nm 8M+1P CMOS process.
This looks like some fun and interesting work for the research group. I wonder what the cost reduction is going to Fourier Transforms, and also whether there are any issues with near-field effects in this scenario.
ReplyDeleteFill factor must be small having strong impact on distance accuracy. Is that right?
ReplyDeletePrinciple wise, I think it's very similar to Fife's multi-aperture camera, correct?
ReplyDeleteSince you mentioned "next generation", I'm wondering what is the current generation 3D technology? It seems to me that people are just trying out different techniques and there is no standards and market is still immature.
@ "what is the current generation 3D technology?"
ReplyDeleteThe current generation is in the products that can be bought right now. First and foremost it's Microsoft Kinect. Also, Softkinetic and PMD, MESA and few other ToF cameras and PointGrey and TYZX stereo cameras are on the market now.
Keith Fife's and few other approaches might, or might not be the "next generation".