基于时空曲线演化的多视频运动对象分割算法

多视频对象由于其运动的复杂性,在分割提取过程中有较大的难度.本文提出了一种基于时空曲线演化的多视频对象自动分割方法,首先根据视频序列帧间(时间域)和帧内(空间域)信息的不同特点,建立基于全局和局部特征的能量模型,并由此导出基于level sets方法的曲线演化方程;然后用视频序列的连继两帧帧差得到初始的视频对象,分别进行时间和空间曲线演化跟踪,提取多个视频对象;当对象因运动而发生相互遮挡现象时,利用基于Bayes最小错误概率决策法则的判断方法,分割遮挡对象和显露对象.实验结果表明,本文提出算法的分割效果在空间准确度上比COST211算法提高30-50%,比最佳的帧差分割算法提高5-10%.

Multiple Video Object Segmentation Based on Spatio-Temporal Curve Evolution

Segmentation of multiple moving object in an image sequence is one of the most challenging problems in image processing due to the complexity of its motion.This paper presents a novel multiple object segmentation algorithm based on spatial-temporal curve evolution.First,According to the dissimilar characteristic of the intra-frame and inter-frame (Spatial and Temporal) information,a joint energy model is proposed with global and local features,thus,a curve evolution equation could be achieved based on the method of level sets.Then,an initial object model is achieved with the difference between two successive frames,multiple objects are tracked and extracted with spatio-temporal curve evolution.Finally,while the occlusion is emerged due to multiple object overlapping motion,the objects could be segmented using Bayes classification for minimum error.The experiment results show that the algorithm is effective.