Interrelationship of Visual Elements of Digital Media Artworks Based on Spectral Graph Theory

This paper first explores the composition of visual elements in modern digital media artworks, extracts the graphical element features of visual elements by improved SIFT algorithm, and classifies and recognizes the graphical elements using by SVM algorithm. Secondly, the extracted and categorized graphical elements are represented by Laplace feature vector correlation spectra in combination with spectral graph theory to study the mutual relationships between the graphical elements. Finally, some graphic elements in modern digital media artworks are used as examples to explore the performance and interrelationship of graphic feature extraction, recognition, and classification. The results show that the vector eigenvalues of spectral graph theory are categorized into [0], (0,100], (100,200], [200, ∞), and the corresponding interrelationships are one-to-one, one-to-many, many-to-one, many-to-many, respectively.