Software A-D Conversion Method for Digital Vision Chips

Abstract

A pixel-level analog-to-digital (A-D) conversion method suitable for digital vision chips is proposed. In this method, each quantization boundary is compared by determining whether the signal level integrated for a certain time has reached a certain voltage, where both the time and voltage are properly controlled by software for each boundary. We derive an algorithm to generate an optimal controlling schedule to realize an arbitrary A-D conversion scale subject to the condition that noise is minimized. Experimental results have shown that images with given A-D characteristics can be obtained with low noise.