Numerous space debris is generated by the growing space activities, which greatly threatens the safety for spaceflight. However, the characteristics of space targets is faint, weak, and small, is different to extract and detect. This paper proposes a space faint target detection method based on noise characterization of Gaussian mixture model (GMM). The method consists of three parts: i) background reconstruction based on denoising diffusion probabilistic models (DDPM); ii) variational inference using GM; iii) extraction and evaluation of faint targets based on the spatial background noise model. The targets with visual magnitudes exceeding 8.5 were detected in space images with a reference target magnitude of 7.5, and the detection results were analyzed based on the target’s radiometric characteristics.