EE Oral Examination: Gordon Wan August 31, 2011

Stanford University PhD Oral Defense - Department of Electrical Engineering

Title: CMOS Image Sensors with Multi-Bucket Pixels for Computational Photography
Speaker: Gordon Wan
Adviser: Mark Horowitz

Date: Wednesday, August 31, 2011
Time: 9:00 AM (Refreshments served at 8:45 AM)
Location: CISX 101 Auditorium

Abstract:
Advancement in digital imaging, in particular image sensor technology, has revolutionized our lives in the last few decades. To mimic the best film, the goals of an image sensor have always been to achieve the most pixels, the smallest cost, the highest sensitivity, the largest signal-to-noise ratio, and etc. Despite that many of these goals compete against each other, image sensors today are amazing in terms of pixel count, sensitivity, low-cost, and read noise. During this sensor improvement, something else has been happening in imaging technology. Recently, a new approach of image creation called computational photography has emerged. In this approach, a picture is no longer taken but rather computed to improve image quality or produce pictures that could not have been taken by traditional cameras. Computing rather than taking a picture, however, is changing the requirements for image sensors.

In this talk, I will present new image sensors with multi-bucket pixels that enable time-multiplexed exposure, an alternative imaging approach. This approach deals nicely with scene motion, and greatly improves high dynamic range imaging, structured light sensing, coded exposure, etc. Using some clever implant design, the new image sensors have successfully incorporated virtual phase charge-coupled device concept into a standard 4-transistor CMOS imager pixel, with small area overhead. Two image sensors with dual and quad in-pixel memories have been designed and fabricated. Pixel sizes of the two sensors represent state-of-the-art for this class of pixel. The technology allows both small cell size, and adjustable charge storage capacity. Using this sensor we have implemented and enhanced some computational photography applications and results agree with theoretical predictions. For computational photography researchers, the sensors will stimulate new software algorithms to be developed by providing a piece of hardware with new functionality.