CSR: Small: Computational Support for Time Domain Continuous Imaging

Most image sensors mimic film, integrating light during an exposure interval and then reading the "latent" image as a complete frame. This model causes a variety of problems in both still and video capture, including reduction in dynamic range, constraints on the exposure interval, and framerate-related motion artifacts. We propose to solve these problems by developing a new approach, time domain continuous imaging (TDCI), in which a continuous waveform is constructed for each sensel describing how the light level at the position changes over time. The ideal implementation of TDCI would use a massively-parallel array of nanocontrollers, each dedicated to sampling the sensel fabricated above it to create and compress a smooth waveform. Under this proposal, we would create and evaluate TDCI implementations using both simulation with conventional sensors and prototypes of new sensors.