Color Pattern Analogy: AI-Assisted Chinese Blue–Green Landscape Painting Restoration

Color analogy is one of the keys to AI-assisted ancient paintings restoration. To harmonize human-AI collaboration in color restoration of Chinese blue-green landscape paintings, we propose a color analogy research paradigm that integrates aesthetic cognition into aesthetic computation by providing data engineering and interactive learning. We conduct research in three phases: data design and scaling for color pattern description, computational experiments for color pattern representation, and color pattern analogy testing. A color pattern analogy model bidirectionally driven by both data engineering and interacted learning is trained. Preliminary tests on two scenarios, restorative colorization prompted by extracted contour and restorative colorization prompted by manuscript, demonstrate that our framework is effective and feasible for color pattern analogy in AI-assisted restoration, thus facilitating the integration of tacit aesthetic knowledge into AI application development.