Multi-focus Image Fusion Via Adaptive Fractional Differential and Guided Filtering
Xiaoling Li,Houjin Chen,Yanfeng Li,Yahui Peng
DOI: https://doi.org/10.1007/s11042-023-16785-w
IF: 2.577
2024-01-01
Multimedia Tools and Applications
Abstract:The goal of multi-focus image fusion is to make an image that is all-in-focus with a finite depth of field. Although multi-focus image fusion algorithms have advanced significantly, it is still difficult to avoid defocused pixels in fused results. To alleviate the problem, a multi-focus image fusion method via adaptive fractional differential and guided filtering is proposed in this paper. Specifically, by combining the guided filtering and the fractional differential, the base and detail layers can be obtained using the designed effective two-scale image decomposition approach. In addition, an image decomposition scheme based on the statistical characteristics concerning the source images is developed, which can retain more detailed information by adaptively adjusting parameters, thereby reducing defocused pixels. The fused all-in-focus images are obtained by merging features from different scales. Experiments demonstrate that compared with existing image fusion algorithms, the promising results on three public datasets are achieved by the proposed method in qualitative evaluation and quantitative measurement, especially in terms of AG, VIF, and SF. On the Lytro dataset, our method has 22%, 34%, and 14% improvement in AG, VIF, and SF compared to the sub-optimal method. On the MFFW dataset, compared with the sub-optimal method, the proposed method has 12%, 14%, and 19% improvement in AG, VIF, and SF, respectively. On the MFI-WHU dataset, the proposed method realizes a performance improvement of 22%, 4%, and 30% in AG, VIF, and SF, compared with the sub-optimal method.