Functional role of input transform in the foveal neural network

The theory and neural network modeling of visual systems of a new type, namely, the foveal type, based on the concept of active vision and using different processing algorithms in the central (foveal) and peripheral vision fields, have been dramatically developed. The long term goal of the present study is the development of a foveal visual sensor based on a nonuniform representation of the vision field. As the first step toward this goal, a simplified neural network model of the FVP has been developed. We investigate the role of the input transform in the FVP by computer simulation and analytical study