Recently, a novel signal-processing tool was proposed, the scattering transform, which uses a cascade of wavelet filters and nonlinear (modulus) operations to build translation-invariant and deformation-stable representations. Despite being aimed at providing a theoretical understanding of deep neural networks, it also shows state-of-the-art performance in image classification. First, we explore its performance for art authentication purposes. We analyze two databases of art objects (postimpressionist paintings and Renaissance drawings) with the goal of determining those authored by van Gogh and Raphael, respectively. To that end, we combine scattering coefficients with several linear classifiers, in particular sparse $\ell_1$-regularized classifiers. Results show that these tools provide excellent performance, superior to state-of-the-art results, and highlight the benefits of sparse classifiers. Second, we show preliminary results of their use for photorealistic rendering. We highlight the limitations of this approach, and the works currently in progress.

Joint work with Yang Wang and Haixia Liu