Optimizing Quantitative and Qualitative Objectives by User-System Cooperative Evolutionary Computation for Image Processing Filter Design

This paper proposes a cooperative optimization method between a system and a user for problems involving quantitative and qualitative optimization criteria. In general Interactive Evolutionary Computation (IEC) models, a system and a user have their own role of evolution, such as individual reproduction and evaluation. In contrast, the proposed method allows them to dynamically switch their roles during the search by using explicit fitness function and case-based user preference prediction. For instance, in the proposed method, the system performs a global search at the beginning, the user then intensifies the search area, and finally the system conducts a local search at the intensified search area. This paper applies the proposed method for an image processing filter design problem that involves both quantitative (filter output accuracy) and qualitative criterion (filter behavior). Experiments have shown that the proposed cooperation method could design filters that are in accordance with user preference and have better performance than filters obtained by Non-IEC search.