Aiming at the difficult problem of color distortion and style adaptation of mural images in the context of cultural and tourism integration, this study proposes a mural style migration model based on the fictionalizable Random Forest algorithm, which is combined with the CycleGAN framework to design a multi-component loss function to achieve a balance between content fidelity and artistic style. The random forest algorithm is improved by joint optimization of probabilistic decision nodes and leaf nodes, path length regularization (PPL) is introduced to control the model complexity, and traceable color profiles are constructed to support mural protection and cultural and creative design. The experimental results show that the proposed model RF-CycleGAN significantly outperforms the comparison algorithm in terms of FID of 175.891 and KID of 0.0233, and the user score reaches 4.08 (out of 5), and the “image appearance” and “style simulation” dimensions in the expert evaluation are 4.08 (out of 5) respectively. The dimensions of “image appearance” and “style simulation” in the expert evaluation score 4.20 and 4.10 respectively. Based on the migrated mural images, we further propose the cultural and creative product design strategy driven by cultural symbols, visual styling and color system, and verify the application value of the model in the living communication of cultural heritage.
