Multi-focus Image Fusion Based on Sparse Features

In order to effectively extract the focused regions from the source images and inhibit the blocking artifacts of the fused image, a novel adaptive block-based image fusion scheme based on sparse features is proposed. The source images are decomposed into principal and sparse matrices by a newly developed robust principal component analysis (RPCA) decomposition. The problem of multi-focus image fusion is transformed into a problem of choosing the sparse features of the sparse matrices to form a feature space. An optimal subdivision of blocks of the sparse matrices is obtained by using a quad tree structure to inhibit the blocking artifacts. The focused regions of the source images are detected by the local sparse feature of the blocks and integrated to construct the resulting fused image. Experimental results show that the proposed scheme can significantly inhibit the blocking artifacts and improve the fusion quality compared to the other existing fusion methods in terms of some objective evaluation indexes, such as structural similarity, mutual information and the edge information transferred from the source images to the fused image.