Implementing the color conspicuity model for visualization

We are developing systems for color selection and placement support, especially for desktop visualization software. One essential part of this system is the conspicuity model, which infers relative, subjective conspicuity between colors. In addition to visibility and discriminability, relative visual weight is an important aspect of any visual design. Previously, we implemented a linear rule-based system for prototype implementation. Recent changes in the prototype allowed us to incorporate simple a neural network which inferred conspicuity much more effectively than the previous model. Testing on actual visualizations showed that the network also predicts ease-of-use for color selection support