Bayesian Nonparametric Intrinsic Image Decomposition^*

Jason Chang, Randi Cabezas, and John W. Fisher III

Abstract. We present a generative, probabilistic model that decomposes an image into reflectance and shading components. The proposed approach uses a Dirichlet process Gaussian mixture model where the mean parameters evolve jointly according to a Gaussian process. In contrast to prior methods, we eliminate the Retinex term and adopt more general smoothness assumptions for the shading image. Markov chain Monte Carlo sampling techniques are used for inference, yielding state-of-the-art results on the MIT Intrinsic Image Dataset.

Keywords: Intrinsic images, Dirichlet process, Gaussian process, MCMC

^*This research was partially supported by the Office of Naval Research Multidisciplinary Research Initiative (MURI) program, award N000141110688, and the Defense Advanced Research Projects Agency, award FA8650-11-1-7154.

LNCS 8692, p. 704 ff.

Full article in PDF | BibTeX

lncs@springer.com
© Springer International Publishing Switzerland 2014