Multi-focus image fusion based on nonsubsampled contourlet transform and residual removal

The goal of multi-focus image fusion is to integrate all focus pixels from the source images into the fused result and simultaneously avoid the introduction of defocused pixels. However, erasing the defocused pixels of the fused image remains a huge challenge. In this paper, a novel multi-focus image fusion method based on residual removal is proposed, which can effectively bridge the gap between the transform domain and spatial domain based methods. Firstly, a structure tensor based fusion rule in nonsubsampled contourlet transform domain is designed, and the initial fused result is obtained. Meanwhile, a new multi-scale threshold correction focusing detection technique in spatial domain is proposed. In this step, all focusing advantages with different scales and focus reliability are taken into account, and then the incomplete decision maps are produced by the pixels with preponderant focus property. Subsequently, the initial decision maps are constructed by the supplement of a novel third-party focus maps. The residual is generated by the initial differences and the residual decision maps. At last, the final fused image is obtained by the subtraction between the initial fused result and the residual. Experimental results demonstrate that the proposed method outperforms some state-of-the-art methods in both quantitative and qualitative evaluations.