Emulation of Salamander Retina with Multi layer Neural Network

Abstract

We utilized a multilayer cellular neural network platform to emulate the functions of salamander retina based on the PIPE model. The template parameters were designed with analytic and embedding methods. By using a FPGA multimedia board for implementation, the emulator completely imitate the function of the PIPE model. The performance of the emulator was evaluated by average SNR and maximum error accumulated in the final image. Both tests confirm the validity of the emulator and the analysis of processing power demonstrated that the emulator can process the input image sequences in real-time with 18-bit precision.