Spatial And Temporal Domain Redundancy Model And Application For Hybrid Coding Stucture

With growing video demands,video coding technology is also rapidly developing.The core contradiction still lies between limited bandwidth and video quality,and how to balance the two is a challenge.Rate-distortion optimization can find the most suitable point among the massive pairs of bitrate and distortion,which is a key technology for addressing this problem.In addition,the Lagrange multiplier can guide decisions such as mode selection and block partitioning in the encoder and is an important parameter.Therefore,this article focuses on rate-distortion optimization and Lagrange multiplier,and the work is divided into three aspects.The first aspect is the block-level rate-distortion optimization algorithm based on residual energy.This algorithm adjusts the Lagrangian multiplier based on the residual energy,which measures the information of the coding block and the motion information of the coding block position.In the LD common test conditions,this algorithm has extremely low time complexity and can achieve certain performance gains.The second aspect is the block-level Lagrangian algorithm based on motion com-pensation entropy.This algorithm estimates the reference value for each coding block to adjust the Lagrangian multiplier.In the LD standard test environment,the BD-RATE performance gain of the Y channel reaches 1.81%.The third aspect is based on the gradient-based block-level Lagrangian multiplier optimization algorithm and the Lagrangian multiplier bitrate difference model.The goal of this algorithm is the I-frame,estimating the complexity of each coding block for spatial redundancy and adjusting the Lagrange multiplier using the model to achieve performance gains.In the AI common test conditions,the algorithm can achieve certain performance gains.The first two algorithms are designed for inter-frame prediction key frames,while the third algorithm is designed for intra-frame prediction key frames.Inter-frame predic-tion and intra-frame prediction cover all predictions within the encoder,and optimizing these two modules can significantly improve the performance of the encoder.These three research works are interconnected and can work together under LD and RA coding struc-tures,enhancing coding efficiency.The three algorithms based on the Lagrange multiplier bitrate difference model proposed in this article were implemented and tested on the VVC standard open source coding platform VTM.According to the experimental results,under standard test environments and their respective coding structures,they can achieve certain subjective and objective performance gains at low time complexity.