Fast CU Partitioning Algorithm In H.266/VVC Based On Entroy

With the new generation of 5th Generation(5G)approaching,information technology in various fields is making rapid progress,as well as computer and network applications constantly innovating.Nowadays,people can obtain information through mobile devices anytime and anywhere,thus the demand for visual information acquisition is also increasing.Therefore,the standard of Versatile Video Coding was announced in 2020 for 4K high-resolution video,which would achieve much better performance in encoding rate compared with the previous High Efficiency Video Coding.Thanks to the introduction of the block division method of the Multitype Tree(MT),the new coding system can match the shape of the divided block more accurately,which greatly improves the accuracy of block division,thereby improving the quality of video coding.But what comes with it is the problem of high calculation complexity,which requires more time to complete the coding process.In order to solve this problem,this paper aims to speed up the coding process and improve the coding efficiency without affecting the visual experience.After describing the complexity of the image content based on concepts originated from relative entropy and information entropy,this approach improves the process of block division in the intra-frame mode and proposes a corresponding optimization algorithm,starting from the similarity of neighboring blocks and the content characteristics of the current block itself.Detailed contents are as follows:(1)Contents in the coding unit(CU)are highly correlated with the coded neighboring block in the same frame.In response,after analyzing the correlation between the current coding block and its neighboring area,using the concepts of relative entropy and information entropy,a fast decision-making algorithm for intra CU based on the similarity of neighboring blocks was proposed.For the current coding block no less than 32×32,the algorithm used the similarity information between the current coding block and the neighboring blocks that had been coded in three directions to skip the MT block division method,and directly selected the method of Quad Tree(QT)to divide the current coding block;Then,the complexity information of the image content itself for the whole image coding block was used to execute the strategy of terminating the large block division early.(2)After analyzing the specific division method of the current coding block,an algorithm for fast decision-making of block division using the standard deviation of one-dimensional information entropy was proposed.Based on the previous algorithm,this new algorithm used multiple sets of one-dimensional information entropy data originated from rows and columns of coding blocks smaller than 32×32,thereby skipping the division methods of horizontal or vertical MT blocks of the current coding block.Then,the one-dimensional entropy standard deviation information of the current coding block in the horizontal and vertical directions was combined to determine the early termination block division strategy of the small block.Both of the algorithms proposed in this research were compared with the original code algorithm of VTM 11.Orcl in simulation experiments.The Bj?ntegaard Delta Bit Rate(BDBR)of the two optimization algorithms proposed in this paper are 1.29%and 1.73%,respectively.Bj?ntegaard Delta Peak Signal-to-Noise Rate(BDPSNR)are-0.052 dB、-0.078 dB,and the coding time saving percentage is 33.11%and 30.46%respectively.The experimental results show that the algorithms proposed in this paper can effectively reduce the computational complexity of block partition selection in H.266/VVC intra prediction process.While maintaining the loss of video coding performance,the algorithms barely affect the sensory visual experience,thus effectively realizing the improvement of video coding efficiency.