A Study On Rate Control In Video Coding

The H.265/HEVC video coding standard,which the Joint Collaborative Team on Video Coding(JCT-VC)introduced in 2013,has dominated a large portion of the market with its superior compression performance.When compared to the previous generation video coding standard,H.264/AVC,H.265/HEVC achieves nearly a 50% reduction in bit rate while maintaining comparable video quality.x265,as a HEVC-compliant encoder,has attracted wide attention due to its open-source nature and real-time encoding speed.Bit rate control plays a crucial role in video coding systems by adjusting coding parameters to generate bitstreams that meet bit rate requirements and achieving higher coding gains by non-uniformly allocating bits to coding units.This thesis focuses on the Average Bit Rate(ABR)bit rate control algorithm of x265 and investigates and analyzes its improvement based on different encoding structures and modes to enhance both the subjective and objective quality of the coded video.The research work of this thesis can be divided into two main points:(1)For the zero-delay encoding structure of x265,this thesis first analyzes the reasons why the Constant Quantization Parameter(CQP)encoding mode achieves better rate-distortion performance than the ABR encoding mode.Then,by referencing the CQP encoding mode,the thesis determines the bit rate ratio between I-frames and P-frames and uses a downsampling complexity estimator to calculate the actual encoded bits and determine the frame-level quantization parameters.A sliding window is designed to adaptively adjust the target bits of encoding frames based on real-time bit rate.Regarding the estimation of quantization parameters at the coding unit level,the thesis first analyzes the additional distortion caused by the current coding unit encoding mode and uses a temporal propagation chain established between two consecutive frames to calculate the temporal influence factor of the additional distortion.The coding unit-level quantization parameter offset relative to the frame-level quantization parameter is calculated based on the temporal influence factor.Compared with x265’s ABR bit rate control algorithm,the optimized algorithm achieves an average bit rate reduction of 5.6%,16.3% and 16.5% for YUV components.(2)For the temporal hierarchical encoding structure of x265,under the CQP encoding mode,the thesis first derives the frame-level temporal influence factor based on the distortion propagation characteristics of coding units.Then,by optimizing the rate-distortion cost of the Group of Picture(GOP)as the objective,the relationship between frame-level Lagrange multiplier and GOP-level Lagrange multiplier is obtained,and the quantization parameter-Lagrange multiplier curve of x265 encoder is re-fitted.The quantization parameter is estimated based on the frame-level Lagrange multiplier.This method achieves an average bit rate reduction of 10.6%,12.7%,and 12.3% for YUV components.Under the ABR encoding mode,a hierarchical rate control model is established.The framelevel temporal influence factor is calculated based on historical information.Then,the bit rate weight of each frame is determined based on the rate control model parameters and temporal influence factor.Finally,bit allocation and quantization parameter estimation are performed based on the bit rate weight and target bit rate.This method achieves an average bit rate reduction of 3.2%,2.9%,and 2.6% for YUV components.