Image stitching technology involves merging several images with overlapping areas into a single high-resolution, wide-angle image that encompasses the information from all the images. Traditional image stitching techniques mostly rely on feature-based approaches. In recent years, with the advancement of deep learning, there has been an emergence of deep learning-based image stitching methods. In this paper, we introduce a novel approach to image stitching through the utilization of an unsupervised learning network. Our proposed network is designed to autonomously learn and execute a sequence of global and local warping operations on input images. Subsequently, the network aligns and stitches the warped images to create a seamless panorama. Compared to conventional algorithms, our method offers enhanced adaptability across diverse scenarios. This adaptability enables our approach to deliver superior stitching outcomes while minimizing the occurrence of ghosting artifacts, thus improving the overall quality of the stitched images.