回归分析建模的自适应SKIP模式预选算法

Adaptive SKIP mode early decision algorithm based on regression analysis modelling

SKIP模式在最佳编码模式中占相当大的比例.SKIP模式预选可以有效减少模式选择的计算复杂度.在现有的SKIP模式预选算法中,该阈值往往不能和视频内容自适应.用回归分析建模的思路,提出了一种内容自适应的SKIP模式预选算法(CASMED),通过一个能够反映当前编码宏块特性的变量来自适应地对阈值模型进行调整.结果证明,该算法在编码性能损耗几乎为零的情况下,大大降低了模式选择算法的复杂度.

SKIP mode accounts for a considerable proportion of optimal coding modes.SKIP mode early decision is an efficient method of alleviating the computation burden of mode decisions.Early decision is usually achieved by threshold judgment based on a certain criterion.In previous research,this threshold was not adaptively refreshed according to the video context.A regression analysis based modeling was employed to construct an accurate model to dynamically estimate the threshold for early decisions.Then,we proposed a context-based adaptive SKIP mode early decision(CASMED) algorithm.Experimental results show that the proposed algorithm considerably reduces the computation complexity with only negligible coding performance loss.