Adaptive multiresolution imager based on FPGAs

This paper describes the architecture of adaptive space-variant imagers with Cartesian structure. Besides their multiresolution outputs, reconfigurable imagers can be adapted for contextual image preprocessing and upgraded for generation of specific data to be processed at the higher level modules of vision systems, making it possible to unload those stages of certain tasks and improving their time response without significant addition of hardware. A synthesizable implementation of these imagers, based on last generation FPGAs adapted to CCD cameras or Active-Pixel digital image CMOS sensors, is also described.