基于图像纹理特征的HEVC帧内编码快速算法

发布时间：2018-05-25 18:18

本文选题：图像处理 + 高效视频编码　；参考：《激光与光电子学进展》2017年07期

【摘要】：为降低高效视频编码(HEVC)的帧内编码复杂度,提出一种基于图像纹理特征的编码单元(CU)划分和预测模式选择算法。利用一种预处理算法来获得当前CU的纹理复杂度和方向。一方面,根据CU的纹理复杂度,该算法自适应地跳过或终止部分CU划分,减少CU深度的遍历时间。纹理复杂度高的CU直接划分成4个子CU,纹理复杂度低的CU将会终止划分。另一方面,根据预测单元的纹理方向,确定相应的候选模式集,通过粗模式决策算法和率失真最优化算法遍历候选模式集选取最优模式。将算法移植到标准食品解码软件HM16.7平台上,实验结果表明,与HM16.7算法相比,编码时间平均减少53.66%,比特率上升0.46%,峰值信噪比下降0.05dB。
[Abstract]:In order to reduce the intra coding complexity of high efficient video coding (HEVC), an image texture feature based coding unit (CU) partition and prediction mode selection algorithm is proposed. A preprocessing algorithm is used to obtain the texture complexity and direction of the current CU. On the one hand, according to the texture complexity of CU, the algorithm adaptively skips or terminates the partial CU partition and reduces the traversal time of CU depth. The CU with high texture complexity is divided into four sub-cubes directly, and the CU with low texture complexity will terminate the partition. On the other hand, according to the texture direction of the prediction unit, the candidate pattern sets are determined, and the candidate pattern sets are traversed by rough pattern decision algorithm and rate-distortion optimization algorithm. The algorithm is transplanted to the standard food decoding software HM16.7 platform. The experimental results show that compared with the HM16.7 algorithm, the coding time is reduced by 53.66, the bit rate is increased by 0.46, and the peak signal-to-noise ratio (PSNR) is reduced by 0.05 dB.
【作者单位】：天津大学精密仪器与光电子工程学院光电信息技术科学教育部重点实验室;
【基金】：国家自然科学基金(61201037)
【分类号】：TN919.81

【相似文献】