A Complexity-Reducing HEVC Intra-Mode Method Based on VGGNet

A Complexity-Reducing HEVC Intra-Mode Method Based on VGGNet

High-efficiency video coding (HEVC) has improved the coding performance by 50% compared with the previous H.264 coding standard. However, it has also introduced an extremely high coding complexity. The quad-tree partition used by the coding unit (CU) is one of the key factors leading to the increase in complexity. Therefore, this paper proposes a CU partition method based on a convolutional neural net-work (CNN). Aiming at the complex recursive calculation of CU partition, an improved VGGNet network structure is proposed to replace the brute-force search strategy, which effectively reduces the computa-tional complexity of intra frame coding. Finally, to enhance the effectiveness of the network model in this paper, the feature pyramid network is added to the CNN model to improve the accuracy of feature extraction. The experimental results show that the proposed method can reduce the intra coding time by 59.71% while maintaining the coding performance.