Intra-prediction Complexity Reduction Using Machine Learning

The new generation of video coding standards, known as VVC, has greatly improved the quality of video coding. The rate-distortion optimization technique must be executed for each division mode while dividing the coding unit (CU). This method increases the complexity of the coding process and time. A VVC intra-complexity prediction algorithm is proposed that utilized the size-adaptive computational neural network. This algorithm aims to create a pre-decision dictionary that considers the various constraints of a given pooling layer and then uses a Convolutional Neural Network model to perform an adaptive decision on the pool size. This method can reduce coding time and avoid unnecessary rate-distortion optimization. Based on the study findings, the suggested technique can decrease the coding time by 43.285% and raise the delta bit rate by a mere 0.575%.