HEVC中运动估计快速算法的研究

发布时间：2018-05-25 07:39

本文选题：视频编码 + 二次函数　；参考：《合肥工业大学》2014年硕士论文

【摘要】：HEVC(高效率视频编码)为继H.264后的最新视频编解码国际标准,由ISO/IEC的运动图像专家组(MPEG)和ITU-T的视频编码专家组(VCEG)组成的联合协作组(JCT-VC)制定的。在相同图像质量的情况下,HEVC的压缩效率比H.264提高一倍以上。然而,高编码压缩率是以很高的计算复杂度为代价的,HEVC标准的计算复杂度约为H.264的2至3倍,因此在实际应用中,要对其算法进行优化以降低其计算复杂度。本论文实现了基于HEVC的运动估计快速算法的优化,论文的工作和研究成果如下：1.在参考软件HM--10.0中,整像素运动估计快速搜索每次都要在搜索中心依次进行步长为1,2,4,8,16,32,64的钻石搜索,本文根据实验统计发现：在搜索中心及步长为1,2中找到最优点的概率最大,因此,本文设置了阈值提前结束搜索,还发现阈值的设置规律：分别在运动几乎静止、平缓、剧烈的视频序列中选择不同的阈值提前结束搜索。2.传统方法用建模的数学函数求出分数像素运动矢量(MY)最优解,然后在最优解附近进行细搜索找到最优匹配点,这样会增加搜索点数,增加计算复杂度。本文提出一种分数像素运动估计快速算法,利用投影为椭圆的两个二次函数得到两个分数像素点,即得到两个分数MV,通过加权这两个点作为最优的分数像素点,即最优MV。实验结果表明在图像质量基本不变,码率略微升高的情况下,该算法的编码时间平均减少了47.44%。3.最后,将本文的整分数像素运动估计快速算法结合起来,实验结果表明,结合算法在视频图像质量和码率变化幅度很小的情况下,编码时间平均减少了58.56%。
[Abstract]:HEVC (High efficiency Video coding) is the latest international standard of video coding and decoding after H. 264. It is made up of ISO/IEC (moving Image expert Group) and ITU-T (Video coding expert Group) (JCT-VCC). The compression efficiency of HEVC is more than double that of H. 264 under the same image quality. However, the computational complexity of the high coding compression ratio is about 2 to 3 times that of H. 264, which is at the cost of high computational complexity. Therefore, in practical applications, its algorithm should be optimized to reduce its computational complexity. This paper realizes the optimization of the fast motion estimation algorithm based on HEVC. The work and research results are as follows: 1. In the reference software HM--10.0, the whole pixel motion estimation fast search must carry on the diamond search in the search center each time in turn, the step size is 1 ~ (2) ~ (2). According to the experimental statistics, the probability of finding the best point in the search center and the step is 1 ~ (2) is the greatest, according to the experiment statistics, it is found that the best one can be found in the search center and the step size is 1 ~ 2. Therefore, this paper sets the threshold to finish search ahead of time, and also finds the law of threshold setting: select different threshold to finish search ahead of time in the moving almost static, gentle and violent video sequences. The traditional method uses the mathematical function of modeling to find the fractional pixel motion vector (MYY) optimal solution, and then carries out a fine search near the optimal solution to find the optimal matching point, which will increase the number of search points and increase the computational complexity. In this paper, a fast fractional pixel motion estimation algorithm is proposed. Two fractional pixels are obtained by using two quadratic functions projected into an ellipse, that is, two fractional MVs are obtained, and these two points are weighted as the optimal fractional pixels, that is, the optimal MVs. The experimental results show that the coding time of the algorithm is reduced by 47.44.3. when the image quality is basically unchanged and the bit rate is slightly increased. Finally, the fast algorithm of integral fractional pixel motion estimation is combined. The experimental results show that the combined algorithm reduces the coding time by 58.56 on average when the video image quality and bit rate change are very small.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】