Existing image outpainting methods still face challenges such as large model parameters and high computational costs, while current boundary padding strategies for convolutional operations tend to disrupt edge continuity. From the perspective of image completion, this paper proposes Pack-based Outpainting and Padding (POP), a unified model that addresses both image outpainting and padding operations. POP reduces computational costs while maintaining high-quality image outpainting performance and explores its effectiveness as convolutional padding operations through semantic segmentation tasks. Specifically, POP employs Pack and Unpack operations to extract image boundaries, which processes boundary regions to reduce computational overhead. Subsequently, the Edge Aware Module (EAM) and Edge Fuse Module (EFM), based on Mamba2 and VSSD components, further extract contextual sequential features and enhance boundary generation quality, thereby achieving high-quality image completion. Furthermore, the lightweight design and reliable outpainting performance of the POP framework enable neural networks with padding operations to achieve more efficient segmentation without introducing excessive computational overhead. Extensive experimental results on five representative datasets in both image outpainting and semantic segmentation domains demonstrate that the POP network surpasses current state-of-the-art image outpainting methods, and segmentation networks utilizing POP modules for padding operations achieve substantial improvements in accuracy.