低复杂度的HEVC帧内编码模式决策算法

发布时间：2018-04-17 03:40

  本文选题：HEVC + 帧内编码　； 参考：《小型微型计算机系统》2017年12期

【摘要】：新一代视频编码标准HEVC虽然显著提升了视频压缩效率,但也大幅增加了视频编码的计算复杂度,其中模式决策部分消耗的编码时间最多.为了降低HEVC编码的计算复杂度,提出一种基于纹理划分特征和方向特征的低复杂度帧内编码模式决策算法.首先根据编码树单元(CTU)的纹理划分特征与最佳编码单元(CU)划分的相关性,通过分析CTU中所有16×16CU的纹理划分特征,自底向上计算不同尺寸CU的纹理划分标识;然后利用这些标识预测当前CTU的深度范围,以及判定是否提前终止CU划分;接着根据预测单元(PU)纹理方向特征与最佳帧内预测模式的相关性,对候选帧内预测模式进行两级选择,以减少进行哈达玛优化的预测模式个数;最后利用哈达玛代价减少进行率失真优化的预测模式个数.实验结果表明,本文算法与HEVC参考模型相比,能够平均降低49.72%的编码时间,而码率只增加0.59%、峰值信噪比仅下降0.04d B,保持了良好的编码率失真性能;与现有的两种模式决策快速算法相比,本文算法进一步降低了约8%和9%的编码时间,并具有相近的编码率失真性能.
[Abstract]:Although the new generation video coding standard HEVC has significantly improved the efficiency of video compression, it also greatly increases the computational complexity of video coding, in which the mode decision part consumes the most coding time.In order to reduce the computational complexity of HEVC coding, a low complexity intra coding mode decision algorithm based on texture partitioning feature and direction feature is proposed.Firstly, according to the correlation between the texture partition feature of the coding tree unit (CTU) and the best coding unit (CU), the texture partition identification of CU of different sizes is calculated from bottom up by analyzing all the texture partition features of 16 脳 16CU in CTU.Then we use these markers to predict the depth range of the current CTU and determine whether to terminate the CU partition in advance; then according to the correlation between the texture direction features of the prediction unit and the best intra prediction mode,In order to reduce the number of prediction modes for Hadamard optimization, the number of prediction modes for rate-distortion optimization is reduced at the cost of Hadamard.Experimental results show that compared with the HEVC reference model, the proposed algorithm can reduce the coding time by 49.72% on average, while the bit rate increases only by 0.59, and the peak SNR decreases only by 0.04 dB, thus maintaining a good coding rate-distortion performance.Compared with the existing two fast algorithms of pattern decision, the proposed algorithm further reduces the coding time by about 8% and 9%, and has similar coding rate-distortion performance.
【作者单位】： 浙江工业大学信息工程学院;杭州电子科技大学计算机学院;浙江大学生物医学工程教育部重点实验室;
【基金】：国家自然科学基金项目(61401398,61471150)资助 浙江省自然科学基金项目(LY17F010013)资助
【分类号】：TN919.81
，

本文编号：1761940