Application of bilateral filtering and Gaussian Mixture modeling for the retrieval of paintings

In this paper we propose a novel approach to the problem of color image indexing and retrieval applied for collections of paintings. The described indexing technique utilizes the Gaussian Mixture Model of the color histogram based on weights provided by the Bilateral Filtering scheme. Thus, the proposed technique considers not only the global distribution of the color image pixels but also takes into account their spatial arrangement. The model parameters serve as signatures which enable fast and efficient retrieval of paintings in large databases. The proposed approach is not only robust to color image distortions introduced by lossy compression artifacts and therefore it is well suited for indexing and retrieval of collections of paintings, but it also provides meaningful results, coherent in their artistic style features.