基于层次B帧预测结构的自适应图像组算法

Adaptive GoP Size Algorithm for Hierarchical B Prediction Structure

与目前普遍应用的IB…BP预测结构相比,在相同参考帧数目前提下,层次B帧预测结构对大部分序列,在相同码率情况下,其图像客观质量PSNR平均提高了1 dB左右。但是,对不同运动特性的图像序列,选择不同图像组大小的层次B帧预测结构编码,编码性能差异较大。基于此,提出了基于层次B帧预测结构的自适应图像组大小编码技术,并给出了具体的图像组大小判断方法和编码过程。实验结果表明,相对于固定图像组大小的层次B帧预测编码技术,对所测序列,本文提出的方法编码性能提高了0.60 dB左右。

With the same reference picture number,hierarchical B prediction structure can achieve about 1 dB gain at the same bit rate in comparison with IBBP prediction structure which are widely used in industry for most sequences.However,for video sequences with various motion characteristics,the coding performances with different GoP sized hierarchical B prediction structure are largely different.Accordingly,an adaptive GoP size algorithm for hierarchical B structure is proposed.In proposed algorithm,the detailed methods for GoP size selection and encoding processes are given.Test results show that,proposed adaptive GoP sized hierarchical B prediction structure can achieve 0.6 dB gain at the same bit rate compared with fixed GoP sized hierarchical B prediction structure.