Abstract: In scenes dominated by indirect lighting, path space filtering methods can achieve higher-quality images with low sample rates, but they primarily focus on spatial correlation while neglecting temporal correlation. Additionally, local path reuse can lead to strong pixel correlation when rendering specular materials. To address this issue, we propose a spatiotemporal reuse path space filtering method. Our method builds on the path graph framework and iteratively improves the radiance estimation in the scene by utilizing samples from both the spatial and temporal domains. Specifically, our method first records information from standard forward path tracing and constructs the path graph using a jittered clustering approach. This clustering is achieved by adding random offsets to the vertex positions recorded during path tracing, effectively eliminating the pixel correlation caused by specular materials. We then introduce a spatiotemporal aggregation operator, which aggregates information within the clusters to refine the radiance estimation of the path graph. Finally, we combine this with the Final Gather operator to further reduce inter-pixel correlation, making the output image compatible with standard Monte Carlo denoisers. In indoor scenes dominated by indirect illumination, experimental results demonstrate that the proposed algorithm achieves an average 42% time reduction compared to path graphs methods during continuous scene rendering, while producing images with specular materials that are more compatible with standard Monte Carlo denoisers.