With the development of the satellite technology, satellite edge computing (SEC) has gained significant attention and become a promising edge computing architecture for 5G-Advanced and 6G, enabling a wide range of on-orbit applications (e.g., remote sensing and cooperative monitoring). However, the heterogeneity of the satellite resources (i.e., computation and communication) makes it challenging to meet the real-time requirements of the applications. In this paper, we propose a novel computation offloading scheme for the SEC networks, which aims to achieve the collaboration among multiple satellites for resource sharing. We further present an offloading decision-making algorithm based on the greedy strategy, which achieves delay optimization under the transmission and computation power constraints. Extensive results based on sensor data generated by satellites show that the proposed scheme can reduce the system delay up to 31%.