Calculating depth image with pixel-parallel processor for a TOF image sensing system

The indirect-TOF principle for the depth image sensing has been widely adopted in the state-of-the-art TOF image sensors. The principle requires arctangent calculation at each pixel to demodulate the depth from the received signal. As the resolution and the frame rate of the TOF image sensor keep growing, the calculation becomes a huge computational burden for conventional serial processing architectures. To solve this problem, we present a pixel-parallel processor to accelerate the arctangent calculation by implementing the CORDIC algorithm in parallel. We come up with a TOF image sensing system comprised of a 256 × 256 TOF image sensor and a 64 × 64 pixel-parallel processor. Under 50-MHz clock, the processor achieves 45.8-Mpixels/s throughput for the depth image calculation, and the induced error is kept within 0.1 cm.