Improvement of S/N ratio of MOS image sensor by neighboring bit correlation method

The spike noise and the fixed pattern noise in MOS type image sensors are quantitatively analyzed, and a new method for noise elimination called the neighboring bit correlation method is proposed. In this method, signals (containing noise) and noise are simultaneously produced from each of neighboring photodiodes, and the difference of these two outputs is taken to eliminate the noise component only. An experimental one-dimensional 512 bit sensor, using two-scanning-circuit scheme, is constructed, and its noise elimination characteristic and photoelectric conversion characteristic are investigated. A high S/N ratio of 30-35 dB is obtained under optimum conditions. The signal output is proportional to the incident illumination intensity, and a gamma value of nearly unity is obtained. The results clearly demonstrate usefulness of the present method for noise elimination in solid-state image sensors.