SoPC implementation of autofocus for infrared imaging apparatus based on retinal vision mechanism

The center problem of common autofocus method by using image processing is to determine the position of best focus of an objective lens through the system output video. The general image processing autofocus algorithms require a focusing window with a suitable size, but this size is difficult to determine. An innovative solution is proposed here according to the retinal vision information-processing function. Un-uniform sampling is introduced by using the retinal space-variant resolution vision mechanism as reference in order to get higher resolution in central part of image and wider view field, making sure that the object located in the fovea region is protruded and the negative effect of the background image is restricted. The sharpness evaluation parameters are further studied for better performance. All the logic circuit above is realized in one FPGA chip of Cyclone II EP2C20 and the SoPC (System-on-a-Programmable- Chip) architecture is adopted to ensure the computing and response speed of the image autofocus. The experimental results indicate that the validity of autofocus algorithm.