Answers are grounded in this paper and cite their sources.
Experimental framework demonstrates 92.7% extraction accuracy across cultural heritage visual symbols, highlighting an efficient approach for digital preservation and creative redesign.
Digital inheritance of intangible cultural heritage requires efficient extraction of visual symbols while preserving morphological details and cultural integrity. Current manual tracing and shallow image-processing approaches are inefficient and may lose symbolic structure or cultural meaning during extraction and transformation. This study proposes an intelligent visual-symbol extraction and creative transformation framework based on deep feature fusion. Representative intangible cultural heritage samples, including paper-cutting, shadow puppetry, and batik, are collected and analyzed from four dimensions: morphology, color, texture, and cultural semantics. A feature database is built to support model training and transformation design. The extraction model integrates convolutional neural networks and generative adversarial networks, with non-stationary fingerprint constraints, multi-window deconvolution decoupling, overlapping masks, delay compensation, and topological consistency constraints to achieve multi-scale feature extraction, interference suppression, and precise segmentation. A Cultural Preservation Index is introduced to evaluate whether extracted symbols retain expert-verified cultural gene templates. Based on extracted symbols, a multidimensional creative transformation strategy is designed, including simplified reconstruction, meaning translation, and scene adaptation. Experiments on 800 visual-symbol samples show an extraction accuracy of 92.7%, significantly higher than single-CNN extraction, while average processing time is reduced to 0.8 s per sample. User recognition of transformed works reaches 86.2%. The framework provides an image signal processing and intelligent visual-design method for digital cultural-resource systems.
No takes yet. Share an insight, caveat, or question.
S. H. Yue (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: