Model For Color Perception System With Learning Capabilities

We present a model for color vision system with learning capabilities. The system adapts to statistical properties of its input. The adaptation is done by utilizing unsupervised learning techniques, as Self-organizing feature maps(1) and vectorial boundary adaptation maps.(2) A color difference reflecting statistical properties of input to the system is defined. The model was tested by using color data with different statistics and two different sets of rhodopsin-based color sensors.