VVC video coding unit division method based on graph neural network

A VVC video coding unit division method based on a graph neural network belongs to the technical field of video coding, and comprises the following steps: S1, collecting video data of VVC coding, and extracting features and label information of coding units; s2, constructing an edge according to a spatial relationship or a time relationship between the CUs; designing a graph attention neural network GNN, and designing an information transmission mechanism capable of simultaneously utilizing node and side information through the graph attention network; s3, a cross entropy loss function is used for solving the six-classification problem of CU division, and then an SGD optimizer is used for training; s4, performing CU division based on the graph neural network; and S5, deploying the trained model to an actual VVC coding system so as to automatically predict and optimize the division mode of the coding units. According to the method, the division problem of the CU is modeled into a multi-classification problem.