The multi-focus image fusion method based on CNN and SR

Abstract

The multi-focus image fusion is a crucial embranchment of image processing, which can obtain better fused consequence from multiple source images. CNN(convolutional neural network)-based and SR(sparse representation)-based image fusion are emerging algorithms in the last decade, and have comprehensive used. So as to gain fused image with more precise and abundant information, this paper proposes a novel multi-focus image fusion method combining CNN and SR. The prevalent SR methods determine the sparse representation vectors after fusion according to ‘max-L1’ rule. But the weighted norm can more accurately reflect the information contained in the source images. Therefore, we choose fused image patches on the basis of the weighted L1-norm, and the weights are got by CNN. Experimental results demonstrate that the proposed method outperforms the existing state-of-the-art methods in terms of both visual perception and objective evaluation metrics.