Figure 5

Information available from different layers and different models. Estimations assume single-step transforms and the same signal-to-noise ratio at every marginal sensor. Left: Mutual information between noise-free LMS and the noisy response at the considered layer (labels at the horizontal axis) for different variations of the model (different color/style). Right: Information transmitted up to the cortical layer by different variations of the model: baseline model (in black), more flexible and more rigid divisive normalization (in red and green, respectively), and totally rigid linear model (in blue). The corresponding versions of these models that neglect chromatic adaptation are represented in the same color but connected by a dashed line