计算资源受限的视频编码多模式决策

发布时间：2019-05-01 06:41

【摘要】：由于视频编码技术趋向于采用越来越复杂的分块模式,多模式决策技术也随之成为一种非常重要的编码技术.多模式决策的优劣不仅会大幅度地影响视频编码的计算消耗,而且也对编码性能的高低起到关键的作用.为使多模式决策在计算能力相差悬殊的平台上都能获得优化的率失真性能,给出一种计算复杂度自适应的优化多模式决策算法.首先,利用视频序列中不同宏块模式间的时空相关性,预测这些宏块多模式决策后的拉格朗日代价和计算复杂度的斜率(Lagrangian cost and complexity slope,简称J-C slope).J-C slope越大,说明在该宏块上的模式决策消耗每单位的计算资源可以获取的率失真收益越多.在计算资源有限的情况下,多模式决策应该按照J-C slope的大小顺序执行,也就是性价比优先的顺序,以便保证计算资源优先分配给率失真收益大的宏块.另外,还通过建立J-C slope阈值与实际计算复杂度的关系模型,设计了一种根据给定计算约束自适应调整计算复杂度的算法.根据实验结果,该方法不仅可以准确地控制多模式决策的计算复杂度,而且还能在不同的计算约束下获得优化的率失真性能.
[Abstract]:As video coding technology tends to adopt more and more complex block patterns, multi-mode decision-making technology has become a very important coding technology. The advantages and disadvantages of multi-mode decision-making will not only greatly affect the computational consumption of video coding, but also play a key role in the coding performance. In order to obtain optimal rate-distortion performance of multi-mode decision-making on the platform with different computing power, an adaptive optimization multi-mode decision-making algorithm with computational complexity is presented. Firstly, using the spatio-temporal correlation among different macroblock patterns in the video sequence, we predict the Lagrangian cost and the slope of computational complexity after these macro-block multi-mode decision-making. The slope (Lagrangian cost and complexity slope, is referred to as the larger the slope of J / C slope). J / C. It is shown that the higher the rate-distortion benefit can be obtained from the consumption of computing resources per unit of the pattern decision-making on the macro-block. In the case of limited computing resources, multi-mode decision-making should be carried out in the order of JC slope, that is, price-to-performance priority order, so as to ensure that computing resources are prioritized to macroblocks with large rate-distortion benefits. In addition, by establishing the relation model between JC slope threshold and practical computational complexity, an algorithm is designed to adjust the computational complexity adaptively according to the given computational constraints. According to the experimental results, this method can not only accurately control the computational complexity of multi-mode decision-making, but also obtain optimized rate-distortion performance under different computational constraints.
【作者单位】：广东金融学院计算机科学与技术系;中山大学软件学院;
【基金】：国家自然科学基金(61173081) 广东省自然科学基金(S2011020001215,S2012040007847)
【分类号】：TN919.81

【参考文献】