基于统计学理论的参数模型运动估计方法

运动估计对视频编码十分重要,基于参数模型的运动估计方法也越来越受到人们的关注,参数模型的选择是该方法的关键。基于此,提出了基于统计学原理的模型选择方法,它以少量的图像数据流为基础,通过参数估计,并分析各近似模型的预测风险和误差,选出最优模型,它最符合预测对象的实际发展变化规律,进而利用该模型对未知对象进行运动估计。试验结果表明,在对实际图像序列进行运动估计时,这种方法是可靠并且实用的。

Parameter model motion estimation using statistical learning theory

Motion estimation plays a very important role in video coding,and people are pay more and more attention to parameter model estimation.Since statistical model selection is the core of that method,this paper proposes a parameter model motion estimation which uses statistical learning theory.This approach could estimate model's parameter from several data sets of image flow,analyze the prediction risk and square error of each possible model in order to get the optimal model that conforms mostly to the real changes of the object.Experimental results demonstrate the feasibility and strength of this motion estimation approach to the real image sequences.