In this study, a low-light video enhancement approach using attention-based deep neural networks is proposed, which consists of encoder, feature propagation module, decoder, look-up table (LUT) generation module, intensity-aware transformation module, and denoising module. Encoder is utilized to extract spatiotemporal information from low-light video frames, feature propagation module is utilized to capture long-term spatiotemporal correlations, decoder is utilized to decode propagation feature maps into intensity maps, LUT generation module is utilized to generate transform tables, intensity-aware transformation module is utilized to perform finer color transformation, and denoising module is utilized to further enhance the quality of video frames. Based on the experimental results obtained in this study, in term of two objective performance metrics and subjective evaluation, the performance of the proposed approach is better than those of four comparison approaches.