Computer Engineering and Applications ›› 2017, Vol. 53 ›› Issue (21): 115-120.DOI: 10.3778/j.issn.1002-8331.1704-0115

Previous Articles     Next Articles

Fast CU partition algorithm for HEVC intra-frame based on random forest classifier

WU Xiaorong1, SHI Zhibin1, LEI Haiwei2, DU Bo1   

  1. 1.School of Computer and Control Engineering, North University of China, Taiyuan 030051, China
    2.Key Laboratory of Instrumentation Science & Dynamic Measurement, Ministry of Education, North University of China, Taiyuan 030051, China
  • Online:2017-11-01 Published:2017-11-15


毋笑蓉1,师智斌1,雷海卫2,杜  博1   

  1. 1.中北大学 计算机与控制工程学院,太原 030051
    2.中北大学 仪器科学与动态测试教育部重点实验室,太原 030051

Abstract: To reduce the coding computational complexity of the quadtree structured Coding Unit(CU) partition process forintra-framein High Efficiency Video Coding(HEVC), a fast CU splitting algorithm based on Random Forest Classifier(RFC) is proposed. The algorithm includes two parts: model off-line training and CU fast coding algorithm. In the process of off-line training, a RFC model is constructed, where the optimal partition result of current CU is utilized as class label(+1, -1), and the contrast, the inverse different moment and the entropy information of current CU are treated as feature vectors. In the process of encoding, characteristic attribute values of current CU are extracted, then, a trained RFC model is used to predict the class label. The experimental results show that the proposed algorithm can save 45.18% coding time on average under the premise of guaranteeing the coding quality compared with HEVC standard algorithm.

Key words: random forest, fast encoding, offline training, CU partition, High Efficiency Video Coding(HEVC)

摘要: 针对HEVC帧内编码中递归式四叉树编码单元(Coding Unit,CU)划分引起的高计算复杂度问题,提出了基于随机森林分类(Random Forest Classifier,RFC)的CU快速划分算法。该算法包括模型离线训练和CU快速编码算法两部分。在模型离线训练中,将CU最佳划分结果(+1,-1)作为分类标签,将当前CU的对比度、逆差矩和熵信息作为特征属性,训练RFC模型。在编码时,提取当前CU的特征属性值,利用训练好的RFC模型快速预测当前CU的划分结果。实验结果表明,该算法与HEVC的标准算法相比,在保证编码质量的前提下,平均可以节约45.18%的编码时间。

关键词: 随机森林, 快速编码, 离线训练, CU划分, HEVC