Spatial Decomposition and Temporal Fusion based Inter Prediction for Learned Video Compression | Synapse