J4 ›› 2013, Vol. 40 ›› Issue (3): 194-200.doi: 10.3969/j.issn.1001-2400.2013.03.029

• 研究论文 • 上一篇    下一篇

利用平滑区域检测的HEVC帧内编码快速算法

蒋洁1;郭宝龙1;莫玮2;焦昶哲1;孙伟1   

  1. (1. 西安电子科技大学 智能控制与图像工程研究所,陕西 西安  710071;
    2. 中国电子技术标准化研究所,北京  100071)
  • 收稿日期:2012-03-12 出版日期:2013-06-20 发布日期:2013-07-29
  • 通讯作者: 蒋洁
  • 作者简介:蒋洁(1983-),女,西安电子科技大学博士研究生,E-mail: jjgirl2008@126.com.
  • 基金资助:

    国家自然科学基金资助项目(61003196,61201290);教育部直属高校基本科研业务费资助项目(K50511040008)

Fast intra coding algorithm using smooth region detection for HEVC

JIANG Jie1;GUO Baolong1;MO Wei2;JIAO Changzhe1;SUN Wei1   

  1. (1. Research Inst. of Intelligent Control & Image Engineering, Xidian Univ., Xi'an  710071, China;
    2. China Electronics Standardization Institute, Beijing  100071, China)
  • Received:2012-03-12 Online:2013-06-20 Published:2013-07-29
  • Contact: JIANG Jie

摘要:

为了降低高性能视频编码(HEVC) 帧内编码的计算复杂度,在其四叉树编码结构的基础上,提出了一种根据平滑区域检测结果跳过子块划分的帧内编码快速算法.首先,通过统计多种不同内容的视频在不同量化参数下最终选取的编码块尺寸的分布,分析了图像平滑特性与编码块尺寸大小的关系;进而根据当前已编码预测单元的编码信息,对当前区域是否属于平滑区域进行检测,根据检测结果自适应地跳过不适合此纹理特性编码块划分的编码过程.在最新的HM4.0平台上的实验结果表明,在相同的重建图像峰值信噪比下,所提算法在低复杂度和高性能标的准测试条件下,编码时间分别减少了24.8%和18.6%,而码率分别只有0.44%和0.12%的细微增加.

关键词: 高性能视频编码, 帧内预测, 四叉树结构, 平滑区域检测, 快速算法

Abstract:

Based on the quadtree structure of the HEVC coding unit, a fast intra coding algorithm that skips splitting subblock is proposed according to the result of the smooth region detection to reduce the intra coding complexity. First, the relationship between the image smoothness and coding unit's size has been analyzed by counting statistically the distribution of corresponding final coding units' sizes selected under different quantization parameters for different video contents. Second, whether the current coding unit belongs to the smooth region is detected according to the coding information on the current coding unit, which can help determine how to adaptively skip the coding process of the coding unit sizes not suitable for the texture features. Experimental results on the latest HM4.0 show that, on average, compared with the default fast approach set in the HEVC reference software, the proposed algorithm reduces the computational complexity by 24.8% and 18.6% for Low Complexity (Loco) and High Efficiency (HE) test sets, respectively, while incurring only 0.44% and 0.12% increment on the total bit rate, at the same peak signal to noise ratio (PSNR), which is of great practical value.

Key words: HEVC, intra prediction, quadtree structure, smooth region detection, fast algorithm

中图分类号: 

  • TP391