A new multi-scale framework for convolutive blind source separation

This paper presents a new multi-scale decomposition algorithm which enables the blind separation of convolutely mixed images. The proposed algorithm uses a wavelet-based transform, called Adaptive Quincunx Lifting Scheme (AQLS), coupled with a geometric demixing algorithm called Deds. The resulting deconvolution process is made up of three steps. In the first step, the convolutely mixed images are decomposed by AQLS. Then, Deds is applied to the more relevant component to unmix the transformed images. The unmixed images are, thereafter, reconstructed using the inverse of the AQLS transform. Experiments carried out on images from various origins show the superiority of the proposed method over many widely used blind deconvolution algorithms.