Visual Perception‐Guided Adaptive Color Mapping Network: Fusing Dynamic Weighted RBF and Deep Residual Optimization

To improve CMYK‐to‐CIELAB forward characterization under nonlinear and boundary‐sensitive conditions, this study proposes an adaptive color mapping network (ACMN) that integrates perceptual‐gated RBF fusion with CNN–MLP residual correction. A CIEDE2000‐guided iterative reweighting strategy is used to improve the perceptual consistency of the interpolation baseline, while a lightweight residual branch captures remaining local nonlinearities in CIELAB space. A continuous gate is further introduced to emphasize elevated‐risk regions without hard regional switching. On the FOGRA39 dataset, ACMN achieves a mean of 0.217 and a maximum of 1.064, with improved upper‐tail error behavior relative to standard RBF. Additional continuity and cross‐dataset results further support the robustness of the proposed method. The proposed framework provides a practical solution for high‐precision and perceptually reliable printer characterization.