Research on the Influence of Color Feature Extraction Methods on Painting Styles Classification

The collection and classification of paintings and artworks digitally has become a critical component of heritage conservation. In different styles and genres of painting, painting techniques and color schemes differ significantly. The existing image classification methods utilize features such as color, outline, and texture to identify paintings, with color being the most prominent. There are a number of specialized extraction calculation methods available in the literature. During the course of this study, an experiment is conducted to compare the effectiveness of existing algorithms for extracting color features from paintings in order to classify them according to their style. This objective is achieved by categorizing and summarizing them. The effects of color space selection, color-position relationship, color and non-color feature combination, and color complexity on classification results are examined based on the minimum closure ellipsoid algorithm. As a result of comparing several sets of experiments, it was found that feature extraction methods that remained true to color-position relationships significantly improved classification results, whereas combining more non-color features or increasing the complexity of color features did not significantly improve classification accuracy. As a result of this study, it is clear that color image-based classification tasks should be focused on discovering color correlations and positional relationships.