Multi-Focus Image Fusion Algorithm Based on Non-Uniform Rectangular Partition and Generative Adversarial Network

Abstract

Based on Non-uniform Rectangular Partition (NURP) and Generative Adversarial Network (GAN), this paper proposes an effective multi-focus image fusion method to generate a full-focus image by combining multi-focus images. Firstly, NURP is applied to left-focus and right-focus images, the size of partitioning grids obtained can be used to judge the fusion pixel to form a rough Fusion Guiding Map (FGM) which will be further optimized by morphological operation and manual adjustment to form an optimized FGM. Then the rough FGM and optimized FGM become the training dataset for the pix2pix GAN. After finishing the training, the trained pix2pix model can be used to optimize any rough FGM from NURP. Finally, the fused pixels are determined according to the FGM to construct the final fused image. The experimental results show that the algorithm improves the visual clarity of the fused image by enhancing the spatial detail of the image and obtains better objective evaluation indicators.