<p>Multi-focus image fusion aims to improve visual performance by integrating critical features from multiple images to create high-quality images with increased resolution and finer details, thereby enhancing overall visual perception. However, most of the current methods use a direct combination of vision transformer and convolution neural networks to extract global and local information from images. Nevertheless, this approach often results in problems, such as high computational demands and blurring of fused image edges. To efficiently capture the overall structural information and edge texture details of images while minimizing computational complexity, we proposed a novel multi-focus image fusion network. It can reduce the computational cost of the model while effectively extracting image features. Specifically, the network consists of two sub-networks, the first sub-network mainly consists of a re-parameterized large kernel convolution network, which extracts both global and local features of the source image simultaneously, thus improving the efficiency of information extraction. The second sub-network facilitates the synthesis of the source image edge texture information mainly through the edge fusion network. In addition, to be able to highlight the edge features of the image, before using the edge fusion network, we enhance the edge features of the source image using the Scharr operator. Our proposed method greatly reduces computational costs while improving the fusion effect. Extensive experiments show that the proposed model achieves state-of-the-art performance in both subjective perception and objective evaluation.</p>