Regional multi-focus image fusion using clarity enhanced image segmentation and sparse representation

To obtain the underlying information from the original multi-focus images and make the fused image clearer, a novel approach based on segmentation of a mix of some clarity enhanced images and the classical sparse representation is proposed. We first use the sparse representation to calculate the relative clarity degree and add it to the original image to construct the clarity enhanced image. Meanwhile, we use the technique of normalized cuts (Ncut) to segment the mix of clarity enhanced images and use the region based method instead of the pixel based method to construct the fused image. The sparse coefficients matrix of fused image is constructed by using the mean-max rule based on the partition results. Finally, the fused image is obtained after inverse transformation. The experimental results demonstrate that the proposed method is a good candidate for multifocus image fusion problems.