Multi-focus image fusion with alternating guided filtering

Due to depth-of-field limitations, existing optical sensors cannot usually obtain a focused image that contains all of the objects in the same scene at different distances. These partially focused images cannot effectively represent the scene and may potentially hinder a complete and accurate understanding of the scene, or even the performance of further computer processes. Multi-focus image fusion can address this problem by fusing multiple images with different focus settings from the same scene into one focused image. This paper proposes a novel multi-focus image fusion method with alternating guided filtering (AGF). First, the source images are decomposed by AGF into base layers containing large-scale variations in intensity and detail layers containing small-scale details. Second, the base and detail layers are fused with saliency maps, which have been filtered using the different layers gradient features. The final fused image is obtained through a recombination of the fused base and detail layers. The experimental results demonstrate that, in terms of visual and quantitative evaluations, the proposed method is capable of representing source images well and demonstrates a significant improvement in fusion quality in comparison with other existing fusion methods.