【摘要】 提出一种基于三维时空小波变换和马尔可夫随机场(MarkovRandomField)模型的多分辨率运动目标分割算法。该算法利用三维时空小波变换对图像序列进行分解得到多分辨率的图像序列,并在此基础上建立多分辨率的马尔可夫随机场模型,构造相应的能量函数。通过条件迭代模型优化算法(IteratedConditionalModes)求解能量函数的最优解,得出标记场,提取出运动目标。实验结果证明,该算法能够很好地消除了单一分辨率的MRF运动检测结果中"空洞"现象,对运动目标分割具有很好的分割效果。更多还原

【Abstract】 In this paper, a moving object segmentation algorithm based on multiresolution MRF using 3D spatio-temporal wavelet transform is proposed. After constructing the image pyramid by using 3D spatio-temporal wavelet transform, energy function of multiresolution MRF model is defined. In order to extract moving object, ICM (Iterated Conditional Modes) is used. Experimental results are provided using Akiyo image sequences. The results show that the proposed algorithm improves the performance to detect moving object with large uniform intensity or slow motion and good segmentation results can be obtained.更多还原