Four-layer cellular neural networks in consideration of color and luminosity

Human´s retina has the capability to identify color and luminosity. The cell identifies color is called a cone cell and identifies luminosity is called a rod cell. The color image processing using CNN was proposed by Roska et al. Additionally, Inoue et al. have used three-layer CNN based on cone cell, and performed edge enhancement. They have confirmed that edge had been detected under three-layer CNN influencing each other. However, the edge of a low luminosity portion is not detected. In this study, we propose four-layer cellular neural networks in consideration of three primary colors of light and luminosity, respectively. In this research, we show some edge detection results and confirm that the proposed CNN is effective compared with the conventional CNN and the existing CNN.