“In response to the high computational complexity and limited extraction of spatiotemporal features of 3D convolutional neural networks in video super-resolution tasks, the research team has designed a lightweight video super-resolution reconstruction network based on hybrid spatiotemporal convolution. The hybrid spatiotemporal convolution module and selective feature fusion system have been established to provide a solution for solving the problem of fast and accurate inference of edge devices.”