Multifocus image fusion using adaptive block compressive sensing by combining spatial frequency

Image fusion is an important branch of the image processing field that makes fuse different information of multiple optical sensors images from the same scene into one complete image. The fused image includes a more dependable and informative description of the scene. With the development of compressive sensing (CS) theory, the compressive domain image fusion method attracts more and more attention. The sampling rate assignment policy for measurement matrix is one of the most important steps in CS and plays a critical role in compression and reconstruction. In this paper, we present a novel multifocus image fusion technique using adaptive sampling rate for block compressive sensing based on textural feature. Firstly, the spatial frequency is utilized to extract the textural features of image blocks. This was then used for adaptive measurement and combining rule. Secondly, the blocks which have large spatial frequency values (e.g., blocks with edges and textures) were assigned high sampling rates. Finally, the combined image was reconstructed with the smooth projected Landweber algorithm. The simulation results show that the proposed method has better performance, in both subjective and objective terms, with respect to the conventional methods.