Super-resolution methods form high-resolution images from low-resolutionimages. In this paper, we develop a new Bayesian nonparametric model forsuper-resolution. Our method uses a beta-Bernoulli process to learn a set ofrecurring visual patterns, called dictionary elements, from the data. Becauseit is nonparametric, the number of elements found is also determined from thedata. We test the results on both benchmark and natural images, comparing withseveral other models from the research literature. We perform large-scale humanevaluation experiments to assess the visual quality of the results. In a firstimplementation, we use Gibbs sampling to approximate the posterior. However,this algorithm is not feasible for large-scale data. To circumvent this, wethen develop an online variational Bayes (VB) algorithm. This algorithm findshigh quality dictionaries in a fraction of the time needed by the Gibbssampler.
展开▼
