Sparse coding algorithm for the visual art style classification

The sparse coding algorithm (SC algorithm) can remove the redundancy and obtain independent features of images. What´s more, the SC algorithm can also distinguish different features. The features extracted by the SC algorithm are helpful to the analysis and classification of the visual paintings´ style. In this paper, the SC algorithm are used to obtain basis functions and sparse coefficients, which are adapted and the sparsest represent of the given paintings Basis functions of the same style are the same and the response of basis functions to paintings of the same style is the sparsest. The kurtosis, which will be high when basis functions and paintings belong to the same style, was used to measure the sparseness and classify different style. The result shows that the SC algorithm can extract the essential characteristics of the images efficiently, and the classification and analysis of the style of visual paintings can be achieved.