PTIFNet: Pseudo-Twin network for multi-focus image fusion

In this paper, we propose a multi-focus image fusion network based on the classical pseudo-twin two-branch network, called PTIFNet, which can preserve the detailed texture structure in the multi-focus image fusion source images and reduce the information loss in the fusion process. The framework is based on the classical twin network, and we input a pair of source images into the corresponding branch for training according to the characteristics of the twin branch network structure, and alternately use the channel attention mechanism and the multiscale convolution module as different stages to extract feature maps. For the purpose of multi-focus image fusion, PTIFNet uses continuous residual blocks to fuse features, and the whole network generates fused images directly without tedious post-processing. The test results of our model obtained are compared quantitatively and qualitatively with 12 existing advanced methods on three test sets, and the effectiveness and generalization of the method are corroborated.