Efficient blind separation of reflection layers with nonparametric transformations

摘要: Superimposed images are very common when taking photos behind glass. We address the reflection separation problem using multiple superimposed photographed in different viewpoints. With viewpoints changing, reflected scenes could contain arbitrarily complicated variations between mixtures, like human's motions or other nonrigid motions. In this article, we propose a moderate hypothesis to tackle scenes' arbitrary as well parametric transformations of transmitted scenes. To rapidly recover high-quality image layers, an Efficient Superimposition Recovering Algorithm (ESRA) by extending framework accelerated gradient method. Our recovering method has good converging performance and is more than 30 times faster state-of-the-art methods. Experimental results on synthetic real world demonstrate that our promising.