【摘要】 视频运动对象的分割技术在运动视觉检测和新的MPEG 4视频编码标准中十分重要 提出了一种运动对象分割算法 该算法采用序列图像帧间差的高阶统计量 (HigherOrderStatistics,HOS)假设检验 ,确定运动对象的位置 ,自动分离运动区域与背景 ;根据三帧序列图像中前后帧差图像灰度边缘重合的部分为中间帧运动对象的边缘来有效地解决运动对象前后帧的遮挡问题 ;采用形态滤波的方法填充分割出的运动对象二值模板中的空洞 ,消除残余噪声及平滑边缘 分析和实验证明 ,该算法需要调整的参数少 ,抗干扰能力强 ,可以高效率地进行运动对象的自动分割 此外 ,该算法具有潜在的并行机制 ,易于实现实时运动图像处理更多还原

【Abstract】 Segmentation of video sequence into independently moving visual objects plays an important role in the motion detection of vision and the new video coding standard MPEG 4 The new sequence segmentation algorithm that extracts moving objects is organized as follows First of all, the higher order statistics(HOS) hypothesis testing of inter frame gray difference is used to automatically separate the motion region from background Secondly, for three consecutive frames, superposition of the gray difference of first and second frame and that of second and third frame highlights the motion edge of the second frame, the covered/uncovered background may be eliminated using such idea Thirdly, using morphologic method to fill and smooth the gained binary mask Experiment testifies that the proposed algorithm is of few parameter, robust to noise, best in result of segmentation and with quick speed更多还原