结合GPM划分的CIIP改进模式探析  

Improved Combined Inter and Intra Prediction with GPM Technology

作　　者：黄炎[1] HUANG Yan(North China University of Technology,Beijing 100144,China)

机构地区：[1]北方工业大学信息学院,北京100144

出　　处：《北京工业职业技术学院学报》2022年第3期35-42,共8页Journal of Beijing Polytechnic College

摘　　要：为了提高CIIP(Combined Inter and Intra Prediction)模式的预测性能,设计改进的MPM(Most Probable Mode)列表,以改善纹理复杂区域中CIIP模式的帧内预测效果。同时,采用GPM(Geometric Partitioning Mode)的划分方式对CIIP模式进行区域划分,划分后的区域通过整数融合矩阵计算最终的CIIP预测值。实验结果表明:与参考模型VTM8.0相比,本文算法Y分量的BD-rate平均降低了0.12%,U,V分量的BD-rate基本不变。In order to improve the prediction performance of CIIP(Combined Inter and Intra Prediction)mode,this paper designs an improved MPM(Most Probable Mode)list to improve the intra prediction effect of CIIP mode in complex texture areas.At the same time,this paper adopts the division method of GPM(Geometric Partitioning Mode)to divide the CIIP mode into regions,and the divided regions use the integer fusion matrix to calculate the final CIIP prediction value.The experimental results show that:compared with the reference model VTM8.0,the BD-rate of this algorithm in the Y components is reduced by 0.12%on average,and the BD-rates in the U components and V components are basically unchanged.

关 键 词：帧间帧内联合预测模式 几何划分 最可能模式列表 整数融合矩阵 

分 类 号：TN919.81[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]