Exploring a unified low rank representation for multi-focus image fusion

Recent years have witnessed a trend that uses image representation models, including sparse (SR), low-rank (LRR) and their variants for multi-focus fusion. Despite the thrilling preliminary results, existing methods conduct fusion patch by patch, leading to insufficient consideration of spatial consistency among patches within local region or an object. As result, not only artifacts are easily introduced fused but also “jagged” frequently arise on boundaries between focused regions de-focused regions, which is inherent problem in these patch-based methods.Aiming address above problems, we propose, this paper,a new method integrating super-pixel clustering unified LRR (ULRR) model. The entire algorithm carried out three steps. In first step, source segmented into few super-pixels with irregular sizes, rather than regular diminish meanwhile preserve objects image. Secondly, clustering-based strategy employed further reduce images. This achieved using proposed ULRR model, imposes constraints onto each cluster.Thisis apparently more reasonable those images complicated scenes. Moreover, Laplacianregularization term incorporated model ensure same cluster. Finally, measure focus defined seek as well thesource via jointly coefficients errors derived from Extensive experiments been conducted results demonstrate superiorities diminishing artifactsin boundary compared state-of-the-art algorithms.