I. Introduction
The resolution of an image is one of the key specifications of image quality, and the image resolution of digital cameras has been increased. This trend is also true in video broadcast applications. In 2002, a 32-megapixel video camera system was demonstrated as a next-generation high-resolution broadcast system [1], which uses three 8-megapixel 2.5-in optical format CCDs [2]. Next, we developed a 1.25-in 8.3-megapixel video CMOS image sensor to reduce both optical size and power consumption in 2003 [3], [4] and the sensor was utilized in a broadcast camera [5]. In this image sensor, a 10-b column-parallel analog-to-digital converter (ADC) was implemented, and the advantages of the CMOS image sensor, such as high-speed and low-power operations, were demonstrated successfully in broadcasting applications. However, further sensitivity improvement and increase of digital resolution were demanded to expand application fields. The major noise source of the sensor was pixel noise caused by a photodiode (PD) reset of a traditional three-transistor pixel. The sensitivity of the broadcasting camera that employed the aforementioned CMOS image sensor was evaluated as 2000-lx F2.8 [5] limited by readout noise. To obtain comparable sensitivity to that of standard CCD broadcasting cameras of 2000-lx F8, the noise level must be reduced to, at least, less than 1/8. Increase of digital resolution was also demanded. Although a 10-b ADC resolution was acceptable in general video image processing, at least 12 b of resolution is required for high-end video cameras, which provides flexibility in the back-end image processing.