Modeling symmetries for stochastic structural recognition

We propose a method for stochastic parsing of images with regular structures exhibiting symmetries, such as facades of buildings. The translational symmetry is represented by an array of elements (windows) that is generated with a stochastic grammar which allows structural exceptions and spatial deviations for individual elements. The reflection symmetry of the elements is automatically inferred as a part of the learning process, where a set of random weak features is boosted into a final mixture. A hierarchical probability model is built for the attributed `words´ generated by the proposed grammar. The image parsing result is then found as the most probable interpretation visited with MCMC sampler which is designed to efficiently explore the space of possible configurations.