Multifocus image fusion and denoising: A variational approach

In this letter we propose a variational approach for concurrent image fusion and denoising of multifocus images, based on error estimation theory and Partial Differential Equations (PDEs). In real world scenarios the assumption that the inputs of an image fusion process contain only useful information, pertinent to the desired fused output, does not hold true more often than not. Thus, the image fusion problem needs to be addressed from a more complex, fusion-denoising point of view, in order to provide a fused result of greater quality. The novelty of our approach consists in defining an image geometry-driven, anisotropic fusion model, numerically expressed using an anisotropy-reinforcing discretization scheme that further increases the anisotropic behavior of the proposed fusion paradigm. The preliminary experimental analysis shows that robust anisotropic denoising can be attained in parallel with efficient image fusion, thus bringing two paramount image processing tasks into complete synergy. One immediate application of the proposed method is fusion of multifocus, noise-corrupted images.

► We propose a variational approach for image fusion with concurrent denoising of multifocus images.
► Novelty of the approach – the development of an image geometry-driven, anisotropic fusion model.
► Fusion model – described using a pixel interpolation scheme + a PDE-based, weighted fusion process.
► Experimental analysis →→ robust denoising can be attained in parallel with efficient image fusion.
► Two paramount image processing tasks are brought into complete synergy.