Multiscale convolutive blind source separation in wavelet transform domain

Abstract

We present a multiresolution approach for blind image separation convolutely mixed. To move in transform domain, we make use of an Adaptive Quincunx Lifting Scheme based on wavelet decomposition followed by a geometric unmixing algorithm. In others words, the mixed signals are decomposed by an adaptive lifting scheme. Then, the unmixing algorithm is applied to the more relevant components. Experiments carried out on images from various origins showed that the proposed method yields better results than many widely used blind image separation algorithms.