A stereo vision system with automatic brightness adaptation

Abstract

This thesis describes the development of an automatic brightness adaptive imaging system for use in stereo vision algorithms implemented for a variety of processing architectures. A 256 x 256 array of wide dynamic range pixels with on-chip A/D converters provides the digital data path for a feedback network which controls the charge integration parameters at each pixel. The first goal of the project was to build a real-time demonstration of the imager with configurable compression functions. Secondly, electronic irising was employed by controlling the global charge integration time based on the average intensity of the image. In addition to electronic-irising, the imaging system employs a linear or a logarithmic compression scheme based on the image data. The controller fits the compression function to the image by comparing the average intensities of many different regions within the image. Finally, a 3-camera stereo-vision system was developed with data transfer to a PC through the PCI bus at 60fps. The imagers are synchronized and controlled based on the center imager's data which allows for consistent object correlation in stereo vision algorithms.