Adaptive Sparse Representation for Analyzing Artistic Style of Paintings

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Inspired by the outstanding performance of sparse representation (SR) in a variety of image/video relevant classification and identification tasks, we propose an adaptive SR method for painting style analysis. Significantly improved over previous SR-based methods, which heavily rely on the comparison of query paintings, our method is able to authenticate or classify a single query painting based on the estimated decision boundary. Specifically, discriminative patches containing the most representative characteristics of the given samples are first extracted via exploiting the statistics of their representations on the discrete cosine transform (DCT) basis. Then, the strategy of adaptive sparsity constraint is enforced to make the dictionary trained on such patches more adaptive to the training samples than via previous SR techniques. Applying the learned dictionary, the query painting can be authenticated if both better denoising performance and higher kurtosis are obtained compared to the baseline estimated via applying the DCT basis; otherwise, it should be denied. Extensive experiments on our dataset comprised of paintings from van Gogh, his contemporaries, the Wacker forgery, and Monet demonstrate the effectiveness of our method.
Funding Information
  • National Surveying and Mapping Geographic Information Public Welfare Industry special funding scientific research project (210600001)
  • National Natural Science Fund of China (NSFC 41101409)
  • Singapore NRF under its IRC@SG Funding Initiative
  • National Basic Research Program of China (2012CB725301)
  • theory and methods of digital conservation for cultural heritage (2012CB725300)
  • IDMPO at the SeSaMe Centre

This publication has 20 references indexed in Scilit: