The goal of the proposal is to develop new image deconvolution and fusion algorithms, which will be applicable in real conditions. We will concentrate mainly on fusion of multifocus images (i.e. images taken with different focal length) and multimodal images (i.e. those taken in different spectral bands such as visible and infrared or by different medical imaging devices). Both these tasks are very important in biomedical imaging. Classification and screening of diatoms (unicellular algae) in images taken from water samples is an examples of a research topic conducted in the host institute that can benefit from the project results, since multifocus fusion provides better visualization of the structural details of diatoms and therefore improves identification process of these unicellular organisms. The review of recent scientific literature shows that there are methods which can fuse images if the degradations are known or if multiple images with space-invariant degradation are acquired. The more general case of space-variant restoration is however common in many application areas but has been an unsolved challenging problem. A satisfactory solution will be of great importance and will have a deep impact on the field of image fusion. In the case of multimodal fusion, development of rigorous formalism for fusion rules would create a solid base for future research in this field and increase applicability of fusion methods. A novel approach with a great potential is to define the fusion rules in the perspective of post-processing tasks performed on the fusion output. It means that the fusion will be subject to a condition that the planned task, such as object classification, is improved.