Research on feature extraction methods for fragmented ceramic artefacts

To address issues such as distorted edge contours and densely intersecting patterns during the extraction of ceramic artifacts, an image processing method combining visual perception enhancement and k-means clustering segmentation is proposed. This approach achieves the complete extraction of fragmented features in ceramic artifacts. Integrating visual perception enhancement, a multi-scale visual perception enhancement equation is established to regulate perceptual contrast and texture across scales, achieving enhanced edge contours and structurally continuous, complete results. By fusing k-means clustering segmentation, a structure-aware adaptive k-means clustering segmentation function is constructed to accomplish precise segmentation of decorative patterns. Results demonstrate that in the task of extracting fragmented features from ceramic artifacts, the PA, DICE, IOU, and Precision metrics reached 92.03%, 89.15%, 85.14%, and 99.38%, respectively. This achieves the complete extraction of fragmented features from ceramic artifacts, bringing breakthrough developments to technological innovation in the field of cultural heritage conservation and restoration.