A New Image-Fusion Technique Based on Blocked Sparse Representation

An image-fusion scheme based on blocked sparse representation is presented in the paper. Firstly, the source images are segmented into patches and then the patches are sparsely represented with learned redundant dictionary. Following that, a salient feature of each sparse coefficient vector is calculated by integrating the sparsity and the \( l^{1} \)-norm of the sparse coefficient vector. Next, the sparse coefficient vectors are fused by adopting the weighted average rule in which the weighted factors are proportional to the salient features of the sparse coefficient vectors. Finally, the fusion image is constructed by the fused coefficient vector with the learned redundant dictionary. Experiments show that the fusion algorithm is effective and superior to the method-based wavelet decomposition.