Design And Implementation Of CU Rapid Partitioning Optimization Algorithm Based On VVC Platform

With the development of communication-related technologies and the emergence of various new entertainment modes,people’s demand for multimedia quality has increased,and the pursuit of higher-quality videos is driving advancements in video coding technology,Currently,the latest generation of coding standards is the Versatile Video Coding(VVC)standard,known for its excellent coding performance.However,its longer encoding time has somewhat limited its popularity in the market.To address this issue,this thesis has conducted extensive research on block partitioning,and the specific work is as follows:1.Proposed a fast algorithm for coding unit(CU)block partitioning based on Discrete Cosine Transform(DCT).Formulas for calculating the squared sum energy and the difference squared sum energy were introduced for texture energy calculation.Different partitioning schemes,including maximum partitioning,minimum partitioning,layered partitioning with averaging,and layered partitioning with weighted values,were proposed based on different CU partitioning methods.Texture complexity(TC)was defined to represent the texture features’ direction and strength of the current CU block,guiding the CU partitioning process.2.Proposed an optimization algorithm for block partitioning based on sub-block correlation.This algorithm is an optimization algorithm built upon the fast algorithm for CU block partitioning based on DCT coefficients.It utilizes a threshold of sub-block correlated texture complexity to determine if the image possesses extremely high texture complexity.For CU blocks with extremely high texture complexity,the sub-CUs inherit the texture features of their parent images without undergoing the DCT transformation step.3.Implemented the above algorithms and designed experiments.Through preliminary experiments to find the optimal texture complexity threshold,comparative experimental results revealed that the layered partitioning scheme with weighted values had better performance and coding complexity considerations when using the difference squared sum energy calculation formula.It showed an average BDBR(BD-bitrate)increment of 0.88%and an average time saving of 35.36%,with the highest time saving reaching 71.1%.The optimization algorithm based on subblock correlation reduced the complexity by an average of 3.21%compared to the original algorithm,with a maximum time complexity reduction of approximately 7.67%and a minimal increase in BDBR of only 0.07%.When compared with other comparative algorithms,for every 1%increase in BDBR,the optimized algorithm saved 40.6%of the time,surpassing the comparative algorithm’s 32.36%.Therefore,it can be concluded that the performance of this algorithm has significantly improved,effectively reducing encoding time and enhancing encoder efficiency.