基于噪声特征参数估计的视频运动区域自动提取

Automatic Extraction of Motion Region in Video Sequence Based on Estimation of Noise Characteristic Parameters

视频运动区域的自动提取是视频对象生成的关键。提出了一种视频运动区域的自动提取方法 ,首先对当前视频帧进行全局运动估计和补偿 ,并和上一帧进行差分运算 ,然后通过估计噪声特征参数自适应地滤除噪声 ,从而提取出运动区域。与传统的差分方法相比 ,该方法具有更强的鲁棒性和精确性。实验结果表明了该方法的有效性。

Automatic extraction of motion region in video sequence is critical to the generation of a video object. This paper presents an algorithm of automatic extraction of motion region in video sequence. Firstly, the current frame in a video sequence is globally motion compensated after the global motion estimation, and the difference operation is performed between the globally motion compensated current frame and the previous frame. Then, through estimating the noise characteristic parameters, noise is filtered adaptively and the motion region is extracted. Compared with the conventional methods, the proposed algorithm is effective with more robustness and accuracy, which has been proved by the experimental results.\;