Taking advantage of data clustering techniques in the multimedia analysis context, the paper entitled “Manifold Learning and Spectral Clustering for Image Phylogeny Forests” describes how to find the image phylogeny forests based on images that inherit content from a single parent. A new approach using manifold learning and spectral clustering is devised to obtain a better representation of the data points distribution and, hence, produce good image clusters.
The approach was first validated and compared with the state-of-art method using quantitative metrics in a controlled scenario. Then the authors used a cross-dataset validation protocol. Finally, the authors were able to evaluate the results in realistic conditions and their behavior in such situations. The results show that the proposed approach is a competitive solution for image phylogeny forests reconstruction, achieving better or equivalent performance when compared to the state-of-the-art approach.
Oikawa, M.A.; Dias, Z.; de Rezende Rocha, A.; Goldenstein, S., “Manifold Learning and Spectral Clustering for Image Phylogeny Forests,” in Information Forensics and Security, IEEE Transactions on , vol.11, no.1, pp.5-18, Jan. 2016 doi: 10.1109/TIFS.2015.2442527